jmrsApril2015CropSmall.jpg

J. Miguel Sanches

Associate Professor
Vice-coordinator for Biomedical Engineering
Bioengineering Department (DBE)
Instituto Superior Técnico, University of Lisbon

IEEE senior member
Institute for Systems and Robotics
IST_A_CMYK_Pos.png
isr_logo_small.png

Home    Projects     Publications

Projects

Medical Imaging

Title: 3D Ultrasound with Kinect (Medical Imaging)
João Sanches

Ultrasound (US) imaging is a useful technology in many fields, including medical, due to it’s portability and low cost. Although 2D data analysis is the standard, results are affected by inaccuracies and subjectivities associated with data acquisition and operator-dependent image selection. 3D data reconstruction could circumvent these limitations, but the costs of the traditional 3D ultrasound systems are usually the bottleneck that prevent its widespread utilization. In this work an alternative is considered where the low cost Kinect sensor is the core of a free-hand 3D ultrasound for carotid atherosclerotic plaque 3D reconstruction.

Objecives: The goal of this work is the development of a free-hand 3D ultrasound system for atherosclerotic plaque 3D reconstruction for computed aided diagnosis (CAD) purposes. The core of the system is the Kynect sensor, from Microsoft.

Description: Carotid Plaque Atherosclerosis accurate and objective characterization is very important for surgical decision making for carotid endarterectomy, a surgical intervention for plaque removal. Recently Activity Index (AI) and Enhanced Activity Index scores were proposed based on clinical data and features extracted from 2D ultrasound images to quantify the risk of stroke. The aim of this work is to design and implement a low cost free-hand 3D ultrasound based on the Kinect sensor, designed to the Microsoft XBox video game console, to  obtain a 3D reconstruction of the plaque and improve the EAI score with 3D information about the plaque.

Title: WebCam based optical Tracker for free-hand US
(10/1/2013)

3D reconstruction of the carotid bifurcation from sequences of nearly parallel cross sections, in a free-hand US basis, is very important for carotid atherosclerosis characterization, quantification and diagnosis. The traditional free hand systems use optical or electromagnetic trackers to measure the position and orientation of the US probe during the acquisition process needed to produce accurate reconstruction results. However, the spatial locators usually used in this type of applications are very expensive because they are general purposes and accurate for a wide spatial range.

In the present application, however, the region of interest is small and the expected probe trajectory is well know, as well as the orientation of the probe, that is expected to present only slight change during the whole course.

Objectives
The goal of this working thesis is to produce an optical spatial locator based on two high-resolution webcams to track the US probe positions and orientations along the acquisition of the image sequences used for 3D reconstruction of the carotid bifurcation and atherosclerotic plaque.

The acquisition protocol imposes tight restrictions to the movement of the probe in order to make the speed as uniform as possible and orientation variations minimal. This constraints should be taken into account in the hardware conception  and algorithm designing in order to maximize the accuracy of the system while keeping it simple and low cost.

Description
The core of the experimental setup is composed by two webcams located near the neck of the patient where the probe is visible during the whole course. A geometric marker placed at the probe is detected in both images of the stero system and the corresponding location and orientation with respect to the 3D real world referential is computed and associated with the acquired US image.

The specific simplified geometry of the problem is taken into account in order to optimize the accuracy of the system, in particular, a Kalman filter is used where the expected dynamics of the probe displacement is used to minimize the measurement errors.

Title: ASL Adpative Optimal Sampling Strategy (AOSS)

Perfusion describes the distribution of nutrients to the tissues by blood flow through the capillary bed and is defined as volume of blood per unit time and per unit volume of tissue. Arterial spin labeling (ASL) magnetic resonance imaging (MRI) techniques offer a non-invasive way of generating perfusion images that are potentially quantitative [1]. They consist on magnetically labeling the water molecules in the blood and then measuring the magnetization of the tissues after a certain time interval, the inversion time (TI). The magnetization difference M(t) as a function of TI in pulsed ASL (PASL) can be described by a standard kinetic model [2], depending on several parameters: perfusion, arterial transit time, bolus time width, blood relaxation rate and a constant related with the tissue relaxation rate. In principle, the magnetization collected at a single TI point is sufficient to obtain a perfusion estimate, provided that the values of the other model parameters are available or can be assumed. However, this is not always the case, particularly in pathological conditions such as cerebrovascular disease. In these cases, it would be possible to estimate perfusion, as well as other unknown parameters, by fitting the PASL model to M(t) data collected at multiple TI points [3]. However, the accuracy of the estimated parameters strongly depends on the distribution of the TI sampling points. Optimal sampling strategies have previously been designed based on the Fisher information matrix optimality criterion for the simultaneous estimation of perfusion and the arterial transit time [4]. However, the optimal time points must be defined a priori and could not be changed adaptively during the acquisition process.

Objectives
Perfusion describes the distribution of nutrients to the tissues by blood flow through the capillary bed and is defined as volume of blood per unit time and per unit volume of tissue. Arterial spin labeling (ASL) magnetic resonance imaging (MRI) techniques offer a non-invasive way of generating perfusion images that are potentially quantitative [1]. They consist on magnetically labeling the water molecules in the blood and then measuring the magnetization of the tissues after a certain time interval, the inversion time (TI). The magnetization difference M(t) as a function of TI in pulsed ASL (PASL) can be described by a standard kinetic model [2], depending on several parameters: perfusion, arterial transit time, bolus time width, blood relaxation rate and a constant related with the tissue relaxation rate. In principle, the magnetization collected at a single TI point is sufficient to obtain a perfusion estimate, provided that the values of the other model parameters are available or can be assumed. However, this is not always the case, particularly in pathological conditions such as cerebrovascular disease. In these cases, it would be possible to estimate perfusion, as well as other unknown parameters, by fitting the PASL model to M(t) data collected at multiple TI points [3]. However, the accuracy of the estimated parameters strongly depends on the distribution of the TI sampling points. Optimal sampling strategies have previously been designed based on the Fisher information matrix optimality criterion for the simultaneous estimation of perfusion and the arterial transit time [4]. However, the optimal time points must be defined a priori and could not be changed adaptively during the acquisition process.

Description
Here, we propose to design, implement and test a different approach where each time sampling is selected in real time according the data previously acquired. Since up to this moment there is no equipment available in our country to run this type of acquisition sequences, we propose to test and validate the algorithm with synthetic data.

References
[1] E T Petersen, I Zimine, Y-C L Ho, and X Golay, Non-invasive measurement of perfusion: a critical review of arterial spin labelling techniques, Br J Radiol, vol. 79, no. 944, pp. 688?701, 2006.
[2] Richard B. Buxton, Lawrence R. Frank, Eric C. Wong, Bettina Siewert, Steven Warach, and Robert R. Edelman, A general kinetic model for quantitative perfusion imaging with arterial spin labeling, Magnetic Resonance in Medicine, vol. 40, no. 3, pp. 383-396, 1998.
[3] P. M. Figueiredo, S. Clare, and P. Jezzard, Quantitative perfusion measurements using pulsed arterial spin labeling: effects of large region-ofinterest analysis., Journal of Magnetic Resonance Imaging, vol. 21, no. 6, pp. 676-682, June 2005.
[4] J. Xie, D. Gallichan, R. N. Gunn, and P. Jezzard, Optimal design of pulsed arterial spin labeling mri experiments., Magnetic Resonance in Medicine, vol. 59, no. 4, pp. 826-834, April 2008.


Biological Imaging and Modeling


Title: E-cadherin 3D spatial distribution characterization for the pathogenicity discrimination of missense mutations from confocal imunofluorescence images
(10/1/2013)
In collaboration with IPATIMUP
Advisers: João Sanches (ISR(IST) and  Raquel Seruca(IPATIMUP/FMUP)

Structural and mechanical properties of the tissues are dependent on the physical linkage between cells. E-cadherin is a key component on this adhesion mechanism and mutations on its coding gene may produce dysfunctional molecules that compromise the cell-cell linkage and increase the risk of cancer.
The homeostatic distribution of E-cadherin within a cell is characterized by a strong and clear concentration at the membrane where it plays its adhesion role. However, for mutated molecules, the trafficking dynamics of E-Cadherin is disturbed and altered distributions of E-cadherin across the cell are observed.
 
Objectives
The goal of this project is to design a method to characterize E-cadherin distribution in the cell with a special emphasis at the junction contacts of cell-cell pairs. Immunofluorescence confocal images will be used. A typical E-cadherin concentration profile along the average linkage of cell pairs is computed representing the distributions for wild type (WT) and mutated molecules.
These characteristic profiles are able to discriminate functional from dys- functional molecules/mutations and can be used for diagnosis purposes. Additional, they can be useful to extract measures with biological meaning such as the expression level of E-cadherin and the relative concentration of the molecule at the membrane aiming to infer the adhesion strength.


 
Description
In epithelia, cell-cell adhesion is achieved by the establishment of homophilic interactions between two adjacent cells. One of the pivotal molecules to attain this homeostatic interaction between cells is the correct localization and function of the transmembrane protein E- cadherin and its interaction with other members of the adhesion complex.
 
In cancer, somatic changes in the expression or function of E-cadherin have been implicated in all steps of tumour progression, including detachment of tumour cells from the primary site, invasion of the adjacent tissue, intravasation into the blood stream, extravasation into distant target organs, and formation of secondary lesions or metastasis. In hereditary forms of gastric cancer germline E-cadherin mutations are causing events. In this lethal cancer disease patients develop highly invasive isolated carcinomas that still pose a very important clinical problem since no effective tools of image screening (endoscopy, MRI and PET) are available yet.
Due to this limitation it is of great importance to develop methods to classify E-cadherin mutants and disclose those who are clinically relevant. We have been focusing on developing in vitro assays to determine the pathogenic relevance of germline E-cadherin mutations in terms of function. However, it is of crucial importance to evaluate the localization and the pattern of E-cadherin expression at the cellular level since aberrant distribution of the protein indicates loss of its normal function: failure in cell-cell adhesion and increased cancer cell invasion. Moreover it is well known that whenever cells lose cell-cell adhesion, the polarity of cells is compromised and the distance between the nucleus in adjacent cells is perturbed. Additionally, this project should investigate the hypothesis of estimating and discriminating the cell cycle phase from DAPI (blue) images of the nucleus. This information could be very useful for an accurate characterization of the E-cadherin distribution at different phases of the cell cycle.


Title: Celular adhesion from fluorescence microscopy images (Characterization of the E-cadherin spatial distribution at a cellular level)
In collaboration with IPATIMUP
Advisers: Raquel Seruca(IPATIMUP), J. Miguel Sanches (ISR/IST)
projec1.jpg

Physical adhesive linkages are crucial for the maintenance of tissue architecture and E-cadherin dependent cell-cell adhesion is central in this process. In cancer, alterations of E-cadherin are very common and significantly associated to loss of cell-cell adhesion and increased invasion potential of cancer cells. In familial forms of gastric cancer (HDGC), E-cadherin mutations may occur at germline level and individuals carrying these mutants are at high risk of developing early-onset (<40 years) invasive gastric cancer without any evidence of symptoms. This is a tragic clinical situation and we urgently need to set-up novel methods to establish the functional relevance of E-cadherin mutations in order to improve the screening and surveillance of these cancer families.

At IPATIMUP, it was developed an in vitro assay to study the effect of E-cadherin mutations in terms of cell invasion, cell-cell adhesion and expression of E-cadherin at cellular level. The later method is based on tagging E-cadherin molecules with fluorescent dyes to observe its distribution in the intra and inter cellular space. In particular, the intensity and distribution of fluorescent E-cadherin at cell membrane, as well as its aberrant presence at the cytoplasm of the cell, is very relevant. This analysis is directly related to the stability of E-cadherin at cell membrane, where it is expected to be in a homeostatic situation establishing a competent cell-cell adhesion complex.

projec2.jpg
Figure 1 - E-Cadherin distribution patterns associated with several mutations.

Objectives:

The goal of this work is to characterize E-cadherin distribution, designing and testing several metrics and features extracted from fluorescence images of the cell with tagged E-cadherin. This analysis will allow the prediction of E-cadherin functional behavior (disease related versus non-disease related). The quantification of E-Cadherin concentration at specific locations, such as the cellular membrane as well as its distribution in the cytoplasm, provides valuable biological information to assess the functional relevance of the mutant forms and to predict the underlying molecular mechanisms associated with its aberrant expression.

Description:

The scope of this work is image processing of fluorescence microscopy images. The distribution of normal E-Cadherin molecules in cells is mainly observed at the membrane, in order to accomplish its role in cell-cell adhesion. In this case, the image intensities at the membrane are stronger than that at the other locations as displayed in Fig.1/WT. Additionally, the distribution of the molecules at the cytoplasm is uniform with a characteristic distribution pattern (textural characterization). On the contrary, the mutated molecules follow a different distribution. Large concentrations in the cytoplasm or lack of E-cadherin signals in some locations are common features for almost all mutations and are associated with loss of cell-cell adhesion. The main goal of this work is to extract morphological and textural features from the microscope images in order to characterize the distribution of E-Cadherin, advance in the underlying molecular mechanisms and predict its role in the adhesion process. The following outputs assume particular relevance for oncologists and biologists dedicated to cancer.

  1. Width of signal intensity in cell to cell contact (Fig.2a)
  2. Number of cell to cell contact points between cells (Fig.2b)
  3. Loss of signal intensity at cell membrane &nbsp;(Fig.2c)
  4. Aberrant cytoplasmic signal intensity (Fig.2c)

projec3.jpg projec6.jpg projec7.jpg

Figure 2 - Biological measures

Due to the small amount of detected fluorescence dots and amplification steps related to the assay, the data collected by these imaging systems exhibit a severe limiting item related to quantification and distribution of the signal per cell and signal-dependent noise. This limitation must be attenuated before functional conclusions can be/are made by the biomedical researcher. In fact, an important task in image processing in this case is the signal quantification and localization within the cell (membrane versus cytoplasm) and denoising, since the data measured by imaging instruments always contain noise and blur. Although being unattainable to build devices that produce data with arbitrary fidelity, it is possible to mathematically reconstruct the underlying images from the corrupted data obtained from real-world instruments. Therefore, that information present but hidden in the data can be revealed with less blur and noise.

Requirements
Image processing knowledge and programming experience, e.g., MatLab or C++. 

Expected results:
Publications in international conferences and journals and a software prototype to process this type of images.


Title: RNA splicing stochastic model and dynamics from spinning disk confocal microscopy images
(In collaboration with IMM/FMUL)

Advisers: José Rino (IMM), Carmo Fonseca (IMM), J. Miguel Sanches (IST)

DNA transcription is one of the central mechanisms involved in the protein synthesis within the cell. This process and in particular the introns removal step, called RNA maturation or splicing, are not completely known and are object of a very active researching work. The improvement and invention of new microscopy image modalities and techniques made it possible impressive advances in this area of biological research. The biologists developed new fluorescent dyes that bind to specific introns. By this, it is possible to observe the splicing process of these introns in real time at each replication site by using the new microscopy image modality, called spinning disk confocal microscopy.
"...The explosive growth in biomedical research using live-cell imaging techniques has been fueled by a combination of events that include dramatic advances in confocal microscopy instrumentation coupled with the introduction of novel ultra-sensitive detectors and continued improvements in the performance of genetically-encoded fluorescent proteins. Acquiring images of localized fluorophores in living cells on the millisecond timescales that reveal intricate biological dynamics presents a host of new challenges, which are far more complex than the traditional issues associated with creating a single high-resolution snapshot of well-stained fixed tissue in a laser scanning confocal microscope. In order to assure high image acquisition speeds of fluorescent proteins and synthetic dyes in live cells with reasonable contrast and minimal photobleaching, microscopes must be able to quickly scan the field of view and record data using detectors with high quantum efficiency. Laser scanning confocal microscopes focus a single beam on the specimen plane to sequentially point-scan a region of interest with spatial filtration of the emission light through a single pinhole that rejects light originating from regions that are out of focus..." (in Zeiss - Introduction to Spinning Disk Confocal Microscopy)
Objectives: The goal of this work is to design a stochastic model to describe the transcription and splicing processes from spinning disk confocal microscopy images. The transcription sites appear in the images as noisy (Poisson distributed)intensity spots where several transcriptions and splicing processes are occurring simultaneously. The final goal is to estimate the number of polymerases and spliceosomes that are acting at each transcription site along the time to extract biological relevant parameters that characterize the dynamic of the process.
 
splicing.png
Description:
DNA transcription is one of the central mechanisms involved in the protein synthesis within the cell. This process and in particular the introns removal step, called RNA maturation or splicing, are not completely known and are object of a very active field of research. The improvement and invention of new microscopy image modalities and techniques made it possible impressive advances on this area of biological research.
The biologists developed new fluorescent dyes that bind to specific introns. By this, it is possible to observe the splicing process of these introns in real time at each replication site by using the new microscopy image modality, called spinning disk confocal microscopy.
These images, however, present a small SNR and are corrupted by a type of multiplicative noise with Poisson distribution due to the huge light amplification need to observe the very weak radiation emitted by the fluorescent dyes.
The main goal is to study the details of the introns removal and degradation processes occurring during the maturation process of the transcribed RNA molecule. The image intensity at each transcription site reflects the number of polymerases operating simultaneously at that location as well as the number of spliceossomes that are maturating the transcribed RNA.
In the scope of this work a stochastic dynamic model should be derived based on the biological knowledge to describe and characterize the dynamics of the transcription and splicing processes. The parameters of the model are estimated from the noisy observation by taking into account the image formation process, namely, its Poisson statistical distribution. Additionally, the transcription site is not static and should be tracked along the 3D field of view along the time at each acquired volume, where several transcription sites are visible.
The work is composed by three main tasks: i) Pre-processing of the images to track the several time change intensity and displacement transcription sites visible in the filed of view, ii) Stochastic model designing based on biological knowledge about the transcription and splicing processes, iii) Estimation of parameters of the model and validation of the method from synthetic data generated with the model, iv) Application of the method to real data.
 
Requirements
Solid mathematical background and programming experience, e.g., MatLab or C++. 

Expected results:
Publications in international conferences and journals and a software prototype to process this type of images.

Neurosurgery


Title: Automatic brain electrical activity registering and analysis for brain functional mapping in Parkinson deep brain stimulation surgery.
(In collaboration with the Neurosurgery department from Hospital Santa Maria, Lisbon)

Advisers: Herculano Carvalho (HSM) and J. Miguel Sanches (IST)

Deep brain stimulation (DBS) is an effective tool in the minimization of the severe symptoms of the Parkinson disease. The accurate position of the electrodes for deep stimulation is a critical procedure with risks. The trajectories of the electrodes need to be careful and accurately planed to not cross vascular and vital structures and to not cause temporary or permanent damages on the brain. Additionally, the final position of the DBS lead is crucial to obtain the desired result that is the minimization of the disabling symptoms associated with the Parkinson disease.
The precise final location of the DBS lead is roughly defined in the pre-surgical preparation procedures. The exact location is only defined during the surgery by making a functional map from the electrical registers obtained along the microelectrodes trajectories. The functional map allows the selection of the depths to test. The intra-operative stimulation test will select the best place to implant DBS lead for symptom suppression without disabling side effects. This procedure is time consuming, mostly made by hand, prone to errors and represents a very important percentage of the total time of the surgery.

Objectives:
The main goal of this work is to design and implement the 5 electrode electrical acquisition system with a wireless interface in order to reduce the amount of hardware in the surgery scenario. The depth control of the microelectrodes is made by the surgeon trough a mechanical device with a screw which is coupled to an encoder that provides a real-time monitoring of depth. The electrical activity in the 5 microelectrodes is automatically acquired and stored at a laptop together with the location (depth) of the microelectrodes.
The software in the laptop generates the electrical activity map along the microelectrodes course in real-time in an incremental basis, that is, as soon as the data is acquired it is immediately used to update the map. In this strategy the neurologist may decide to inspect locations previously observed if more accuracy is needed.

Description:
Deep brain stimulation (DBS) is a surgical procedure used to treat a variety of disabling neurological symptoms—most commonly the debilitating symptoms of Parkinson’s disease (PD), such as tremor, rigidity, bradykinesia and gait disturbance. The procedure is also used to treat essential tremor, dystonia and other movement disorders. At present, the procedure is used only for patients whose symptoms cannot be adequately controlled with medication.
DBS uses a surgically implanted, battery-operated medical device called a neurostimulator— a dual-channel device capable of bilateral brain stimulation similar to a heart pacemaker. The device delivers controlled electrical pulse to stimulate targeted areas in the brain that control movement, blocking the abnormal activity of deep brain nuclei that cause the movement disorder.
Before the procedure, a neurosurgeon uses magnetic resonance imaging (MRI) and stereotatic computed tomography (CT) scan to identify and locate the exact target and plan a surgical trajectory to reach a nucleus within the brain. A neuronavigation system is used to do the surgical plan.
At surgical time, some surgeons may use microelectrode recording—which involves a small bipolar wire that monitors the activity of nerve cells in the target area—to more specifically identify the precise brain target that will be stimulated. Then is performed the intra-operative stimulation test to verify the symptoms improvement and absence of significant side effects. Finally is implanted the DBS lead. Generally, these targets are the thalamus, subthalamic nucleus(STN), or globus pallidus.
 
projec1.jpg projec2.jpg
 
The DBS system consists of three components: the lead, the extension, and the neurostimulator. The DBS lead (also called a definitive electrode)—a thin quadripolar, insulated wire—is inserted through a small opening in the skull (burr hole) and implanted at the selected nucleus. The tip of the lead is positioned within the targeted brain area, and can deliver stimulation either by one pole (monopolar) or by a combination of poles (bipolar).
The extension is an insulated wire that is passed under the skin of the head, neck, toward torax, connecting the lead to the neurostimulator. The neurostimulator (the "battery pack") is the third component and is usually implanted subcutaneously in subclavicular thoracic region. In some cases it may be implanted under skin at the lower chest or abdomen.
Once the system is in place, programmed electrical impulses are sent from the neurostimulator up along the extension wire and the lead and into the brain. These impulses interfere with and block the electrical signals that cause the movement disorder."
Adapted from “ NINDS Deep Brain Stimulation for Parkinson's Disease Information Page (http://www.ninds.nih.gov/disorders/deep_brain_stimulation/deep_brain_stimulation.htm) “
 
 
projec3.jpg projec4.jpg projec5.gif projec6.jpg
 
 
This work is composed by two main tasks of hardware and software. The goal of the first task is to implement the acquisition setup of the signals collected from the microelectrodes implanted in the brain by using a wireless system in order to minimize the wires at the surgery environment that disturb the procedure and more important, improve the signal to noise ratio of the acquired signals by reducing the artifacts and interferences that corrupt them. The depth of the microelectrode is controlled by the surgeon but measured by an encoder. The information of depth provided by the encoder is sent to the computed together with the electrical registers from the microelectrodes by the wireless communication channel, monitoring their position along the planned trajectory. The second task is the real time analysis of the acquired signal at the computer in order to compute a functional map along the microelectrodes course. This signal analysis is done according the state of the art methods described in the literature and the results and control of the process is done throught an user interface application in C++ running on the computer. 

Expected results: The expected results are an experimental prototype, a patent and publications in international top conferences and journals.
Requirements:  Solid mathematical background and programming experience, e.g., MatLab or C++. 

Sub-topics
Title:  Functional brain mapping for neurosurgery planning from scalp and intracranial electroencephalography data
Title: CT and/or MRI co-register with US for computer assisted surgery (CAS) in neuro-surgery guidance
(In collaboration with the Neurosurgery department from Hospital Santa Maria, Lisbon)

Advisers: Herculano Carvalho (HSM) and J. Miguel Sanches (IST)

Requirements:  Solid mathematical background and programming experience, e.g., MatLab or C++. 
Title: Surface brain shift deformation tracking with structured light in brain surgery
(In collaboration with the Neurosurgery department from Hospital Santa Maria, Lisbon)

Advisers: Herculano Carvalho (HSM) and J. Miguel Sanches (IST)

Brain shift is a common effect observed in open brain surgeries where the slow leakage of cerebral fluid leads to a deformation of the whole brain and of the cortex in particular. This deformation leads to a misalignment with the MRI/CT images acquired during the pre-surgical phase of the intervention which difficult the work of the surgery team and will result in inaccuracies in surgical procedures that may cause severe damages on surrounding regions.

Objectives: The goal of this work is to design and implement a real time compensation method of brain shift based on images of the exposed part of the brain where a structured pattern of light is projected. This pattern, reflecting the morphological shape of the cortex is registered with the pre-surgical MRI/CT volumes in order to adapt them to the slowly evolved deformation associated with brain shift.
Description:
 
Requirements:  Solid mathematical background and programming experience, e.g., MatLab or C++. 
Bibliography
Tuhin K. Sinha, Benoit M. Dawant, Valerie Duay, David M. Cash, Robert J. Weil, Reid C. Thompson, Kyle D. Weaver, and Michael I. Miga, A Method to Track Cortical Surface Deformations Using a Laser Range Scanner, IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 24, NO. 6, JUNE 2005.



Brain functional Imaging and Computer Interface

Title: Scalp EEG Synchronization and desynchronization quantification in the analytical signals framework
(10/1/2013 - CoAdviser: Profª Patrícia Figueiredo)

The EEG is one of the most common techniques for brain activity assessment and it is very well established in the clinical practice for diagnostic purposes. Recently, it has received additional attention and developments in the scope of emergent techniques like Brain Computer Interfaces (BCI) or simultaneous EEG-fMRI. In these modalities, detection of event-related potentials (ERP) and associated event-related desynchronization / synchronization (ERD/ERS) are central goals in several scientific and clinical scopes, e.g., detection of motor movement or epileptic activity. The most commonly used method for assessing ERD/ERS consists on the evaluation of EEG power changes upon the event onset in relation to the baseline, reflecting increased/decreased synchronization.

Objectives
The goal of this project is the designing of a new method for ERD/ERS estimation based on measuring spatial variations of the EEG phase. A continuous phase map is estimated from the discrete set of EEG traces by using the Hilbert transform in the analytical signals framework. The synchronization at each arbitrary continuous location in the scalp space is then computed as the magnitude of the phase gradient at that location. The interpolation/fitting algorithm is a central issue of the project due the highly time and space variability of the interpolated signals.

Description
The electroencephalogram (EEG) is one of the commonest techniques used in the study of neuronal activity and to assist the diagnosis of brain disorders. Event-related potentials (ERPs) are defined as the changes of the activity of neuronal populations elicited by sensory stimuli [1,2]. The usual way to detect ERP is by averaging the responses over the several trials. This approach admits that ERPs consist of a pattern time and phase-locked in relation to the stimuli [2]. Moreover, changes in the ongoing EEG activity elicited by specific cognitive or pathological events often occur in specific frequencies bands, hence a time-frequency analysis of the signals is desirable.

During a motor task, like voluntary movement, desynchronization phenomena occur in the upper alpha and low beta frequency bands across the motor cortex. Such desynchronization is observed predominantly in the hemisphere contralateral to the movement close to the scalp projection of the hand area (electrode C3 or C4, for right or left hand movements, respectively) prior to movement onset. But immediately after movement onset, the desynchronization is bilateral and nearly symmetrical. It means that the neuronal groups oscillate in a frequency range during the movement and after that they enter into phase-locking over a period of time. There is evidence that these phase desynchronizations and synchronizations between neural signals are reflected in the EEG in terms of the so-called event-related desynchronization/synchronization (ERD/ERS).
This phenomenon is usually quantified as the relative change in spectral power elicited by the event in relation to the baseline period, assuming that an increase / decrease of the EEG power in a certain frequency band corresponds to the synchronization / desynchronization of the underlying neuronal population’s activity, respectively. Alternatively, ERS/ERD could also be assessed based on the phase differences across EEG channels, which are thought to reflect different degrees of synchronization between the underlying neuronal populations. The Phase-Locking Factor (PLF), which measures the phase difference between two channel’s signals, has previously been used as a measure of synchronization between the two signals. A PLF close to 1 means that the two signals are synchronized (with the same phase), while a PLF close to 0 means that the signals are desynchronized (with different phases).

In this project, a novel method to assess ERD/ERS based on the continuous variation of the phase across EEG channels will be designed. The topographical distribution of the EEG channels on the scalp is represented in a 2D matrix and the Hilbert transform is used to obtain the phase of the EEG signal recorded from each channel at each time point. The ERD/ERS is then computed as the modulus of the phase gradient along both directions. The proposed methodology was applied to the EEG data recorded from a healthy volunteer performing an index finger movement. The results were compared with the conventional even-related spectral power ERD/ERS measure.




Biofeedback

Biofeedback is a recent and general framework where physiological measurements are used do provide the patient with objective indicators about his own state. The goal is to help the subjects to learn and train procedures to self-control some of their brain and physiological states, such as muscle relaxation.
Biofeedback systems based on electromyography (muscle activity) are the most traditional ones and the first used in clinical practice to help patients in muscle relaxation.
Recently, biofeedback systems based on Heart Rate Variability (HRV) have received a lot of attention because they allow a direct monitoring of the autonomic nervous system (ANS) that depends, among other factors, on the central nervous system (CNS).
By this, biofeedback systems seem particularly suitable to help patients with several psychological and psychiatric disorders such as anxiety and panic, sleep disorders, Depression, Stress, Fatigue and Attention deficit / hyperactivity disorder  ADHD (http://www.centerforadd-az.com/)
The huge increasing on the computational power of the smart-phones allows, today, the implementation of highly portable, complex and sophisticated medical applications impossible some years ago.

Title: A biofeedback system for insomnia
João Sanches and Teresa Paiva

Objectives
The goal of this work is to study and implement a biofeedback system and training program for insomnia based on MAPSA (Mobile Acquisition Platform for Sleep Assessment), a mobile platform based on smart-phone under development in our group.

Description
MAPSA is being developed for long term physiological data monitoring to be used in the diagnosis of sleep disorders. This platform, highly portable based on smart-phone, is suitable for a wide range of applications, such as biofeedback ones.
The objective of this work is to develop a biofeedback application in MAPSA based on HRV and other physiological and environment data to help patients with insomina. Besides the engineering component, the work contains a clinical component where patients from the Centro de Elecroencefalografia e Neurofisilogia Cl_nica (CENC) will collaborate for validation and testing purposes.

Requirements:  Solid mathematical background and programming experience, e.g., MatLab, Java or C++. 
Expected results: Publications in international conferences and journals and a software prototype to process this type of images.

Title: A biofeedback system for panic and anxiety
João Sanches

Objectives
The goal of this work is to study and implement a biofeedback system and training program for help subjects with Anxiety and Panic Attacks based on MAPSA (Mobile Acquisition Platform for Sleep Assessment), a mobile platform based on smart-phone under development in our group.

Description
MAPSA is being developed for long term physiological data monitoring. This platform, highly portable based on smart-phone, is suitable for a wide range of applications, such as biofeedback ones.
The objective of this work is to develop a biofeedback application in MAPSA based on HRV and other physiological and environment data to help subjects suffering from Anxiety and Panic Attacks.
Requirements: Background on mathematics and physiology.
Outputs: Patent and publications on international journal and conference.
Location: ISR/IST and Institute of physiology at the IMM/FMUL

Title: Biofeedback in MAPSA

Individuals exhibit voluntary yet unwittingly control over brain activation in a daily basis. Nevertheless, the extent to which an individual can learn to directly and consciously control the activation of specific brain regions is not yet known. Biofeedback is a procedure supported by machinery that translates physiological processes into audio or visual real-time signals, with the goal of controlling the nervous system activity.

The development of portable devices and systems for biofeedback purposes is an active field of research with unquestionable importance in a medical perspective, e.g., in psychiatry, psicology or sleep diagnosis and treatment, the main focus of our research.

Objectives
The inclusion of a EEG acquisition system, with at least two channels, in the Mobile Acquisition Platform for Sleep Assessment (MAPSA) and the implementation in the platform of a Biofeedback application to illustrate the concept and show the usefulness and feasibility of the platform for these type of medical applications.

Description
The goal of this work is to incorporate an EEG signal acquisition functionality in the Mobile Acquisition Platform for Sleep Assessment (MAPSA). This platform, based on a mobile phone, has been developed in the last four years in the scope of several MSc thesis for the acquisition of physiological,  clinical and patient data aiming at the monitoring and diagnosis of sleep disorders.

By including the EEG functionality in the platform it is expected to improve its power form a diagnosis point of view but also the increasing of  its application range, namely, for Biofeedback purposes.

Therefore, its is expected, as an output of this work, besides the EEG acquisition functionality, the implementation of a Biofeedback application on the platform, using the acquired EEG and ECG data, to be used in the scope of sleep disorders therapy.

The results of this work should be submitted for publication in an international journal and conference. It is also expected the submission of a patent proposal.

Emotional and Cognitive quantification

Title: Real time stress quantification from physiological and behavioral data

Stress assessment from physiological and behavioral data is very important in several practical applications, such as vehicle driving. In most of the cases the existing methods and systems are cumbersome and non-portable, which prevents its use in real life conditions.

Stress and attention manifest on the autonomic nervous system activity that can be indirectly "observed" trough some physiological variables that can be easily acquired, such as the ECG and Galvanic Skin Response (GSR).

Description
This thesis is about stress and attention quantification and detection from heart rate variability (HRV) and GSR data. As shown in [1], physiological and behavioral data have the ability of sleep staging and hypnogram estimation.  In the scope of this work a prototype, based on an arduino, will be designed and constructed to acquire in real time, ECG, actigraphy and GSR data from the left wrist and shoulder of the subject.

The data, acquired with this bracelet-type system, highly portable and non-cumbersome, is transmitted to a smartphone to be stored and processed in real time. The software application, running in the smartphone, detects stress, attention deficit or drowsiness and takes countermeasures to prevent catastrophic events, e.g. car crashing.

Objectives
This work is composed by two main tasks: 1) Hardware design and assembling and 2) algorithm design and implementation. In the first step an Arduino microprocessor/microcontroller based system is used to acquire ECG, Actigraphy and GSR data at the non-dominant wrist and shoulder of the subject and sent that data to a smartphone. In the second step, algorithms running at the smartphone analyze the data to quantify the levels of stress, attention and sleep.  The software should generate countermeasures, such as visual or audio stimulus, to induce appropriated levels of attention, decrease stress and prevent drowsiness.

Requirements:
Mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences and a prototype.

Location:
ISR/IST and CENC

References
[1] Domingues, A.; Paiva, T.; Sanches, J., "Hypnogram and Sleep Parameter Computation from Activity and Cardiovascular data," Biomedical Engineering, IEEE Transactions on , vol.PP, no.99, pp.1,1, doi: 10.1109/TBME.2014.2301462, 2014
Title: Dreams, Emotional contents and Brain computer interfaces

Dreaming is a highly complex process not yet completely understood. One of its known functions is the consolidation of lived experiences during the awakening state. Some of these consolidation tasks contain motor components that can be detected and processed under the classical framework of motor imagery in brain computer interface (BCI) systems.

Description

In this work, EEG data collected during sleep, in the scope of polysomnographic exams, are analyzed and processed with well-established techniques of brain motor imagery, in order to detect types of dreamed movements and actions. A brain computer interface system is trained during the awake state with well established motor paradigms aiming at detecting broad classes of movements, such as, left or right limb movements during sleep.

Objectives

The objectives of this work is to design a brain computer interface system, typically used during the awakening state in several interaction tasks, such as with handicapped subjects, in the scope of sleep and dreaming studies. The physiological foundations of the problem need to be studied and experimentally validated in order to prove the hypothesis that motor imagery is able to detect imagined movements during sleep and dreaming.

Requirements:
Mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences and a prototype of the detection system.

Location:
ISR/IST and CENC
Title: Dream contents estimation from polysomnographic data

Dreams are by far one of the most unknown phenomena in terms of standard physiological data.
Dreams, mostly those occurring in REM sleep, have important emotional components. Emotions in dreams can vary between negative and positive modalities, being the negative ones more frequent.
 
In awake states emotions can be detected by changes in ECG namely due to tachycardia measured from the ECG waveform. The autonomic control is different in awake and REM sleep. Furthermore, REM sleep is associated with frequency changes in HR.  However, it is not known whether these changes are associated with specific emotional states or other dream components.

Description
In this work, dreams collected during REM sleep are characterized in a psychological basis from dream reports collected in normal subjects in the sleep laboratory during REM awakenings.
Additionally, since the most important component of the HRV is related with the autonomic nervous system, other measures of its activity are also acquired such as galvanic skin response (GSR). The time constant differences on these two responses need to be taken into account.

The dream reports are stored together with neurophysiological data that include ECG and GSR information, for offline processing and analysis with statistical signal processing techniques and machine learning tools.

Objectives
The main goals of this work are: 1) Correlate the psychological emotional reports with the physiological/autonomic data and 2) find specificities of this correlation.

A set of labels, representing broad classes of emotional states during dreaming are defined and correlated with physiological data using data mining and machine learning tools, in order to infer the emotional contents during sleep without having to awake the patients.

Requirements:
Mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences.

Location:
ISR/IST and CENC
Title: Dream emotional states, Time event detection and Autonomic activity by HRV/ECG waveform analysis and GSR data

Up to the moment it is not known how to identify dream states from physiological data or even if a subject is dreaming or not. In spite of this it is known that dreams contents are correlated with some changes in EEG frequencies. However the specific correlation is not known.

Description
In this work, dreams collected during REM sleep are characterized in a psychological basis from dream reports collected from normal subjects in the sleep laboratory during REM awakenings.
Additionally, since the most important component of the heart rate variability (HRV) is related with the autonomic nervous system, other measures of its activity are also acquired such as galvanic skin response (GSR).

The dream reports are stored together with neurophysiological data, which include ECG information, for offline processing and analysis with statistical signal processing techniques and machine learning tools.

Objectives
The main goal of this work is to identify the exact moments of dream events and the classification of the corresponding contents. This implies: 1) using variables describing the observations, e.g., EEG, EOG, EMG and ECG and 2) apply classification techniques of temporal series to identify dream states. 3) The ground truth is in this case the existence or not of dream reported during the forced awakening from REM sleep.

The ultimate goal is detecting the presence of  dreaming states and the corresponding onset time events to automatically annotate long duration polysomnographic data.

Requirements:
Mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences.

Location:
ISR/IST and CENC

Epilepsy

Title: Characterization and detection of epileptic seizures from physiological and behavioural data

Advisors: João Sanches (ISR/IST) Carla Bentes (Neurologia, HSM/FMUL) and Rita Peralta (Neurologia, HSM/FMUL)

Description
:
Detection and characterization of epileptic seizures is a very important goal for an accurate knowledge of the disease, in order to adopt the appropriated therapeutic measures and strategies. Moreover, the automatic detection of epileptic seizures is also very important for sake of patient safety. Epileptic seizures, especially if consciousness impairment or generalization occurs, may be associated with severe traumatic injuries. Furthermore, epilepsy-related causes of death account for 40% of mortality in these patients and include accidents during the epileptic attack but also sudden unexpected death episodes (SUDEP).

Detection of seizures is accurate in laboratorial and hospitalar conditions where several data modalities, e.g. EEG, ECG and video, are acquired with highly complex and sophisticated equipment and analyzed by experts.

These exams, however, are expensive, time consuming, highly cumbersome and intrusive for the patients, preventing its use in ambulatory conditions during normal day life conditions. Additionally, the difficulty of monitoring the patient for long periods during his quotidian life also prevents the characterization of uncommon or patients unrecognized seizures. 

Objectives:
The objective of this work is to design, implement and validate a portable system to monitor and detect epileptic seizures, especially with motor phenomenology, during wakefulness and    sleep, based on physiological and behavioral data.

The main sources of data to achieve the proposed goal are the ECG and Actigraphy.
The ECG will be used to perform Heart Rate Variability (HRV) analysis to estimate the autonomic activity.  A seizure can present with autonomic symptoms and autonomic dysfunction due to epilepsy can contribute to sudden unexpected death episodes.

Actigraphy data, acquired from the non dominant wrist of the patient, in ambulatory conditions and from the four limbs at the Hospital, contain useful information about the motor activity of the patient.
In this work, a set of discriminative features are extracted from the HRV and Actigraphy data aiming at detecting seizures and generate alarms and automatic annotations that will help the medical staff in the detection and characterization of the disease. Special attention will be paid to seizures during sleep, frequently unrecognized by the patient in his daily life. Part of the work will investigate the discriminative power of actigraphy data acquired from an actigraph sensor located in the patient bed especially tuned to detect generalizes seizures, only from movement patterns.

Requirements:
Solid mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences and a prototype of the detection system.

Location:
ISR/IST

Title
: Epileptic seizure detection in children during sleep

Objectives:
The objective of this project is to create and test algorithms for detecting seizures related to the movement of actigraphy data starting symptoms.

The objective of this project is twofold: to study and characterize typical movement patterns during seizures in children, from a fundamental research perspective, and to design and implementing of a prototype system to alert parents or medical team of epileptic seizures onset.

Description:
Epileptic seizures with movement disorders are potentially dangerous in children during sleep because they can cause injuries as a result of uncontrolled movements during the crises.
However, movements during sleep are normal and a simple movement detector to detect seizures is useless.

Actigraphy is the natural choice to this project. Actigraphy signals, obtained from 3D accelerometers usually placed at the non-dominant wrist of the subject, are able to discriminate normal from disturbed movements where global movement features, not  associated with specific tasks, can be captured. Additionaly, actigraph sensors are noninvasive and do not involve any experimental apparatus that can interfere with the normal sleep process of children.

So the goal of this project is to use actigraph data obtained from childrens during normal sleep and during epileptic seizures and apply statistical signal processing tecniques to compute relevant features and use machine learning algorithms to detect epileptic sizure events.

Requirements
:
Solid mathematical background and programming experience in MatLab.

Results:
Papers in international journals and conferences and a prototype of the detection system.

Location:
ISR/IST

Physiology

Title: Quantificação da actividade do SNA a partir do ritmo cardíaco e dilatação das pupilas em ambiente de condução automóvel
João Sanches (ISR/IST), 2013

Objectives
O objectivo deste projecto é o de estudar, desenvolver e implementar métodos para quantificar a actividade do Sistema Nervoso Autónomo e da atenção a apartir da variabilidade cardíaca, da dilatação das pupilas e de um canal de EEG, usando métodos de processamento de imagens e processamento de sinais fisiológicos para efeitos de monitorização em tempo real do estado de atenção e concentração em ambiente de codução automóvel.

Description
A actividade do sistema nervoso autónomo  depende fortemente do sistema nervoso central e portanto é influenciado por estados cerebrais, tais como atenção, sono e relaxamento, e emocionais, tais como ansiedade e pânico, A avaliação em tempo real da actividade do SNA é impotante em diversas aplicações relacionadas com o sono, disfunções psiquiátricas ou em aplicações específicas como sejam a avaliação emocional de condutores em tempo real para detectar sonolência ou falta de atenção durante a condução.

O objectivo deste projecto é essencialmente o de desenvolver os algoritmos e montar o aparato experimental que permita criar um protótipo para esta última aplicação, isto é: através da aquisição do ECG, de um canal de EEG e da análise das pupilas do condutor, detectar estados cerebrais não apropriados para a condução e gerar os alarmes apropriados para evitar acidentes.
Title: Desenho, teste e validação de um modelo do sistema cardio-vascular e a sua interacção com o sistema nervoso autónomo: baro and chemo reflexes
João Sanches (ISR/IST)  e Isabel Rocha (IF/FMUL), 2012

Objectives
Desenvolvimento de um modelo, já parcialmente implementado emMatLab/Simulink, que reproduz o comportamento da pressão arterial e seus determinantes ( frequência cardíaca, débito cardíaco, resistência vascular periférica, volume sistólico, amplitude e frequência dos movimentos respiratórios)  num teste de stress respiratório ( respiração profunda)
O objectivo deste modelo é compreender a contribuição individual e integrada destes componentes no controlo integrado cardio-respiratório.
Pretende-se também que este modelo possa ser utilizado durante as aulas de fisiologia como simulador da regulação integrada cardiorrespiratória.

Description
O sistema nervoso autónomo, em conjunto com os sistema endócrino e imunitário, contribui para a manutenção da homeodinamia  controlando as funções viscerais instante a instante. Neste sentido, torna-se fundamental a sua avaliação fidedigna com o objectivo de melhor entender os mecanismos fisiológicos e, também, os fisiopatologicos subjacentes a patologias cardiovasculares que cursem com alterações respiratórias.
Existem conjuntos de manobras provocativas do sistema nervoso autónomo que permitem uma avaliação da regulação do SNA, de uma forma não invasiva e reproductivel. Conceptualmente, o teste  da respiração profunda permite avaliar o comportamento do reflexo quimioreceptor e a sua intervenção na regulação integrada cardiorrespiratória em Indivíduos normais e em doentes.
Title: Desenho, teste e validação de um modelo do sistema cardio-vascular e a sua interacção com o sistema nervoso autónomo: reflexos cardíacos
João Sanches (ISR/IST)  e Isabel Rocha (IF/FMUL), 2012

Objectives
Desenvolvimento de um modelo, já parcialmente implementado emMatLab/Simulink, que reproduz o comportamento da pressão arterial e seus determinantes ( frequência cardíaca, débito cardíaco, resistência vascular periférica, inotropismo, volume sistolico)  num teste de stress ortostático  ( teste de ortostatismo passivo/teste de tilt)
O objectivo deste modelo é compreender a contribuição individual e integrada destes componentes no controlo da pressão arterial.
Pretende-se também que este modelo possa ser utilizado durante as aulas de fisiologia como simulador da regulação do sistema cardio-vascular pelo sistema nervoso autónomo.

Description
O sistema nervoso autónomo, em conjunto com os sistema endócrino e imunitário, contribui para a manutenção da homeodinamia  controlando as funções viscerais instante a instante. Neste sentido, torna-se fundamental a sua avaliação fidedigna com o objectivo de melhor entender os mecanismos fisiológicos e, também, os fisiopatologicos subjacentes a patologias cardiovasculares.
Existem conjuntos de manobras provocativas do sistema nervoso autónomo que permitem uma avaliação da regulação do SNA, de uma forma não invasiva e reproductivel. Conceptualmente, o teste de ortostatismo passivo permite avaliar o comportamento do reflexo baroreceptor e a sua intervenção na regulação da pressão arterial em Indivíduos normais e em doentes.
Title: Desenho, teste e validação de um modelo do sistema cardio-vascular e a sua interacção com o sistema nervoso autónomo: reflexo baroreceptor
João Sanches (ISR/IST)  e Isabel Rocha (IF/FMUL), 2012

Objectives
Desenvolvimento de um modelo, já parcialmente implementado emMatLab/Simulink, que reproduz o comportamento da pressão arterial e seus determinantes ( frequência cardíaca, débito cardíaco, resistência vascular periférica, volume sistólico, inotropismo)  num teste de stress ortostático  ( teste de ortostatismo passivo/teste de tilt)
O objectivo deste modelo é compreender a contribuição individual e integrada destes componentes no controlo da pressão arterial.
Pretende-se também que este modelo possa ser utilizado durante as aulas de fisiologia como simulador da regulação do sistema cardio-vascular pelo sistema nervoso autónomo.
 
Description
O sistema nervoso autónomo, em conjunto com os sistema endócrino e imunitário, contribui para a manutenção da homeodinamia  controlando as funções viscerais instante a instante. Neste sentido, torna-se fundamental a sua avaliação fidedigna com o objectivo de melhor entender os mecanismos fisiológicos e, também, os fisiopatologicos subjacentes a patologias cardiovasculares.
Existem conjuntos de manobras provocativas do sistema nervoso autónomo que permitem uma avaliação da regulação do SNA, de uma forma não invasiva e reproductivel. Conceptualmente, o teste de ortostatismo passivo permite avaliar o comportamento do reflexo baroreceptor e a sua intervenção na regulação da pressão arterial em Indivíduos normais e em doentes.
Title: A Mathematical model of the Cardiovascular System in the scope of heart coherence

The cardio vascular system (CVS) promotes the blood flow across the body in  order to properly deliver oxygen and nutrients to the cells at every organ and tissue. This feedback system operates in a close-loop basis in order to adapt the blood flow and pressure in the large and small vessels according the external and internal specific conditions that are mainly related with the interaction of the subject with the surrounding environment.
These physiological parameters, flow and pressure, are controlled by the autonomic nervous system (ANS) by acting in the heart and vessels. The rate and contractility of the heart and the section of the small and medium vessels are the main parameters controlled by the ANS in order to keep the constancy of the blood pressure (BP), called Homeostasis.
The control of this system, performed by the ANS, is mainly performed in great extend at involuntary  and unconscious level. However it is known that the central nervous system (CNS) is able to influence several parameters of the CVS. Additionally, there are several techniques reported in the literature that are able to change some parameters of the CVS, most of them based on voluntary procedures related with respiration.
Heart Coherence seems to be a resonant condition of the CVS where regularity patterns and magnitude of the Heart Rate Variability (HRV) are maximized. The specific set point is associated with health conditions and positive brain states. More important, accurate and realistic mathematical models to describe the CVS can be of great help to understand and diagnosis some dysautonomia disorders, such as Vasovagal Syncope (VS) and Postural Orthostatic Position Syndrome (POTS) or other very common disease, some times with unknown etiology, like  hypertension.

Objectives
The goal of this work is to derive a physiological based mathematical model to describe the CVS and simulate the respiration effects on it. The model, with parameters tuned with information obtained from physiology literature,  should be able to reproduce the physiological measures of heart rate and mean arterial pressure (MAP) recorded during the Hewing protocol with the tilt table usually used to assess the performance of the ANS and detect dysautonomia diseases, such as, VS and POTS. The model should also be able to reproduce Heart Coherence with a proper respiration excitation.
 projec1.gif
Figure - Tilt table used in the Hewing protocol for ANS assessment

Description
Heart, vessels and receptors (baro and chemio) are the main efferent components of the CVS controlled by the ANS, located at the brain stem. This system can be modeled under the Control Theory framework as a canonical feedback architecture. The control pathway is composed by the two branches of the ANS, parasympathetic (vagus) and sympathetic fibers that code the control signal by mean of the firing rate intensity.
The model, based on physiological information, will be simulated with the graphical programming language Simulink from MalLab and it is expected to improve the previous Master thesis [1].
 
 projec8.jpg
Figure - Baroreflex simulink model

The main goal of this work is to improve the model derived in [1] (displayed in Figure above) by including the effect of respiration in order to simulate and reproduce the heart coherence, a pocessess involved in relaxation and emotions self-regulation.

Requirements: Background on mathematics and physiology.

Outputs: Patent and publications on international journal and conference.

Location: ISR/IST and Institute of physiology at the IMM/FMUL

References:
[1] José Santos, A Baroreflex Control Model Using Head-Up Tilt Test, Master Thesis at IST and Inst. Fisiologia IMM/FMUL 2008.


Title: Modeling of the Haemodynamic Response Function based on Physiological Principles

The metabolic activity of brain cells may be indirectly measured by several means, namely, by the BOLD signal obtained by using functional MRI (Magnetic Resonance Imaging) or directly by using invasive methods. The metabolic activity of these cells is related with the brain activity as response of external or internal stimulus and induces changes on several physiologic variables such as blood flow and oxygenated / non oxygenated hemoglobin ratio.
The relation of the blood flow and oxygenated hemoglobin concentration with brain activity is a complex process depending on several physiologic mechanisms, which are not completely understood. On the other hand, this knowledge is crucial for several purposes, such as, robust indirect detection of activated region in the functional MRI scope.

Objectives
The goal of this work is to deeply study the physiologic mechanisms involved in the hemodynamic response of the tissues (namely in the brain) to design realistic theoretical sound and physiologic based mathematically models for this response. Validate the model indirectly with fMRI and directly by invasive methods in mouse.

Description
This work is the continuation of the Master Thesis [Rita Gafaniz, 2010]  aiming to derive a mathematical model to describe the relation between electrical neuronal activity and metabolism. This is a basic and central issue in several scopes such as functional MRI. However, it is also related with the basic processes involved in sleep.

Requirements:  Solid mathematical background and programming experience, e.g., MatLab or C++. 

References
[1] COUPLING OF NITRIC OXIDE TRANSPORT AND BLOOD FLOW: A COMPUTATIONAL MODELING APPROACH, A Thesis Submitted to the Faculty of Drexel University by Xuewen Chen in partial fulfillment of the requirements for the degree of Doctor of Philosophy, March 2007.
[2] Timing of potential and metabolic brain energy, Jakob Korf and Jan Bert Gramsbergen, Journal of Neurochemistry, 2007, 103, 1697–1708.
[3] Dynamic Modeling for Modulation of ATP Concentration at the Endothelial Surface by Viscous Shear Flow, Kairong Qin, Zhe Xu, Hui Zhang, Cheng Xiang, Shuzhi Sam Ge and Zonglai Jiang, Proceedings of the 2006 IEEE International Symposium on Intelligent Control Munich, Germany, October 4-6, 2006.
[4] Vasoconstrictor and Vasodilator Effects of Adenosine in the Mouse Kidney due to Preferential Activation of A1 or A2 Adenosine Receptors, P. B. Hansen, S. Hashimoto, M. Oppermann, Y. Huang, J. P. Briggs, and J. Schnermann,  THE JOURNAL OF PHARMACOLOGY AND EXPERIMENTAL THERAPEUTICS Vol. 315, No. 3, JPET 315:1150–1157, 2005.
[5] Discovery of Some of the Biological Effects of Nitric Oxide and its Role in Cell Signaling, Ferid Murad, Bioscience Reports, Vol. 24, Nos. 4/5, August/October 2004 ( (c) 2005).
[6] Nitric Oxide: Involvement in the Nervous Control of Cardiovascular Function, L. N. Shapoval, Neurophysiology, Vol. 36, Nos. 5/6, 2004.
[7] Sameer A. Sheth, Masahito Nemoto, Michael Guiou, Melissa Walker, Nader Pouratian, and Arthur W. Toga, Linear and Nonlinear Relationships between Neuronal Activity, Oxygen Metabolism, and Hemodynamic Responses, Neuron, Vol. 42, 347–355, April 22, 2004.
[8] Brain Function and Neurophysiological Correlates of Signals Used in Functional Neuroimaging, Martin Lauritzen and Lorenz Gold, The Journal of Neuroscience, May 15, 2003 • 23(10):3972–3980.
[9] MOLECULAR-BIOLOGICAL PROBLEMS OF DRUG DESIGN AND MECHANISM OF DRUG ACTION: METABOLISM OF L-ARGININE (A REVIEW), V. G. Granik, Pharmaceutical Chemistry Journal Vol. 37, No. 3, 2003.
[10] A Model of the Coupling between Brain Electrical Activity, Metabolism, and Hemodynamics: Application to the Interpretation of Functional Neuroimaging, Agn_s Aubert and Robert Costalat, NeuroImage 17, 1162–1181 (2002).
[11] The neural basis of the blood-oxygen-level-dependent functional magnetic resonance imaging signal, Nikos K. Logothetis, Phil. Trans. R. Soc. Lond. B (2002) 357, 1003–1037.
[12] How well do we understand the neural origins of the fMRI BOLD signal?, Owen J. Arthurs and Simon Boniface, TRENDS in Neurosciences Vol.25 No.1 January 2002.

Title: Muscle activation and strength: a dynamic model
 
João Sanches(IST) and Mamede de Carvalho(FMUL)
 
Objectives
The goal of this work is the derivation of a theorectical dynamic model for motor unit recruitment on mild muscular contraction, imvolving frequency and regularity of the motor unit firing . The validation of the model will be done with synthetic and real data. The aim is to investigate the hypothesis that single motor unit recruitment variation is dependent on the normal function of the cortico-spinal tract. This work involves laboratorial work at the Physiology Institute in the FMUL to acquire real data and measure neuronal activity, needed to validate the model.

Description
The strength of the muscle depends on the number and frequency of the recruited motor unit. On slight contraction the different motor units are easily recognized. Mild contration activates a limited number of motor units which fired at a frequency of about 7 Hz. The inter-discharge interval of single motor units activation shows some variation. Based on previous exploratory investigation, this variation is decreased in patients with corticospinal tract lesion. We plan to establish a recording set up to register motor units recruitment on mild contraction in controls and patients. The goal of this work is derive a mathematical model to describe this process where the motor units are modeled in a stochastic framework. This model can be used to explain and help in the diagnosis of certain neurological disorders where motor function is disturbed. In the scope of this work, laboratorial experiments should be designed to acquired data at the Institute of Physiology – FMUL, used to tune and validate the model.

Title: Estimation of the haemodynamic response to epileptic activity in EEG-fMRI data

Advisers:
João Sanches  and Patrícia Figueiredo (IST)

The recent technical advances in the simultaneous acquisition of the electron-encephalogram (EEG) with functional magnetic resonance imaging (fMRI) have made it possible to localize epileptogenic brain networks with good spatial resolution, through the identification of the haemodynamic correlates of (inter-) ictal electrical discharges. One of the major limitations of this technique has been the uncertainty regarding the exact shape of the haemodynamic response function (HRF), which may hinder its detection sensitivity. In fact, the commonly used canonical HRF is most likely altered in epileptogenic brain regions and it would therefore be desirable to employ data analysis methods that allow HRF estimation simultaneously with activity detection.

Objectives
Estimation of the HRF to inter-ictal and ictal discharges from EEG-fMRI data in focal epilepsy.Estimation of the HRF to inter-ictal and ictal discharges from EEG-fMRI data in focal epilepsy.

Description
The work will consist on the implementation of appropriate algorithms for the analysis of simultaneous EEG-fMRI data, collected from epileptic patients during both inter-ictal and ictal discharges. In particular, a new method will be tested for the estimation of the HRF, simultaneously with activity detection, based on a Bayesian approach. The results will be compared with alternative methods used in the literature.

References
  1. LeVan P, Tyvaert L, Moeller F, Gotman J. (2010) Independent component analysis reveals dynamic ictal BOLD responses in EEG-fMRI data from focal epilepsy patients. Neuroimage. 49(1):366-78.
  2. David Afonso, Jo_o Sanches and Martin H. Lauterbach (2008), Joint Bayesian detection of brain activated regions and local HRF estimation in functional MRI, ICASSP.