|
Credit:
1 course unit
Elective course
Catalog Description:
This course will provide practical education in engineering technologies used to monitor and modulate the nervous system and their translation into clinical devices. Fundamental concepts in neurosignals, hardware and software will be reinforced by practical examples and in-depth study of three neurodevice platforms over the course of the semester: (1) localization of epileptic networks with intracranial electrodes, and modulation of these circuits with responsive brain stimulation (2) localization and stimulation of thalamic nuclei for treatment of movement disorders (e.g. Parkinson's disease), (3) systems for evoked-potential driven computer-guided communication for quadriplegic patients. Basic background in neurosignals will be provided, spanning scales from single neurons to large-scale field potentials, and across modalities from electrophysiology to optical and chemical recording. Algorithms for extracting, classifying, and modulating neurosignals will be covered, along with strategies for reducing them to practice on implantable computational platforms. Finally, some appreciation for hardware implemented in clinical systems will be given, along with their limitations and major design considerations. By the end of the course students will be able to design and implement a scaled-down brain-computer interface device in computer software simulations, and understand basic concepts involved in its implementation and approval.
The course is geared to advanced undergraduates and graduate students interested in understanding the basics of implantable neuro-devices, their design, practical implementation, approval, and use. Reading will cover the basics of neuro signal recording, analysis, algorithms for controlling therapy and fundamental concepts governing clinical implementation, approval, and use. The focus of the course will be on lectures and homework assignments that build incrementally towards culmination in a complete Brain-Computer Interface (BCI) design. Guest lecturers and demonstrations will supplement regular lectures.
Prerequisites:
BE301 (signals and systems) or equivalent, computer programming experience, preferably in MATLAB (e.g., as used in the BE labs, BE209/210/310). Some basic neuroscience background [e.g., BIOL215, BE305, BE520, INS (neuroscience) core course], or independent study in neuroscience, is required. This requirement may be waived based upon practical experience on a case by case basis by the instructor.
Grading:
50% Homework & Assignments
20% Midterm
30% Final Project (examples at the end of the syllabus)
Textbook(s) and/or
other Required Material:
Bulk-pack of articles and book chapters including:
1. Chapters from: Pedley and Ebersole, Eds., Current Practice of Clinical Electroencephalography, Lippincott, Williams and Wilkins.
2. Chapter: Echauz J, Wong S, Smart O, Gardner A, Worrell G, and Litt B: Quantitative methods for tracking seizure generation in epileptic networks Computational Neuroscience in Epilepsy. Soltesz I and Staley K (eds.). Elsevier, 2008.
3. S Jensen, G Molnar, J Giftakis, W Santa, R Jensen, D Carlson, M Lent and T Denison. Information, Energy, and Entropy: Design Principles for Adaptive, Therapeutic Modulation of Neural Circuits. Plenary Talk Summary, European Solid-State Circuits Conference [ESSCIRC], Sept 2008.
4. Kossoff E, Ritzl E, Politsky J, Murro A, Smith J, Duckrow R, Spencer D, Bergey G. Effect of an External Responsive Neurostimulator on Seizures and Electrographic Discharges during Subdural Electrode Monitoring. Epilepsia. 2004 Dec; 45(12):1560-7.
5. Wong S, Danish S, Jagi J, and Baltuch G. Guiding Electrode Placement For Deep Brain Stimulation By Fuzzy Cluster Multi-Unit Activity. IEEE Transactions on Biomedical Engineering, in press.
6. Schalk G. Brain-computer Symbiosis. J. Neural Eng. 5 (2008) P1-P15.
7. Lewicki, M. S. A review of methods for spike sorting: the detection and classification of neural action potentials. Network 9, R53-78 (1998).
8. Buzsaki, G. Large-scale recording of neuronal ensembles. Nat Neurosci 7, 446-51 (2004).
9. Krusienski D, Sellers EW, Cabestaing F, Bayoudh S, McFarland D, Vaughan TM, and Wolpaw JR. A comparison of classification techniques for the P300 Speller. J. Neural Eng. 3 (2006) 299-305.
10. Viventi J, Maus D, Litt B. Evolution of Technology for Monitoring Brain Networks. Review Manuscript, in submission.
Tools:
This course is designed around two core libraries: (1) a library of data collected from actual patients and devices, and (2) a library of algorithm routines that the students will use during the course to analyze the data archive. Homework will utilize these tools, and develop increasing proficiency gradually over the semester, culminating in their final projects.
Topics Covered:
Week 1:
Introduction; Basics of Neurosignals: Signal Generators, dipoles, cells and circuits. Multi-scale recordings: EEG, evoked potentials, field potentials, units.
Homework: Basic patterns in neurosignals.
Week 2:
Introduction to the feature library and human data archive. Basic algorithms: Computing basic features and classifying output using
the library and archive.
Guest Speaker: Javier Echauz, Ph.D., author of library tools.
Homework: Downloading features from the library and simple processing assignment.
Week 3:
Event detection and basic classifiers: signal averaging, clustering, post- processing.
Guest Speaker: Andrew Gardner, Ph.D., Momentics Corp.
Homework: Separate two neural signals from a group of brief recording segments.
Week 4:
BCI Hardware I: types, systems, biocompatibility
*Introduce final projects
Guest Speaker: Tim Denison, Ph.D., Medtronic, Inc.
Homework: Design specifications for BCI hardware given specific components.
Week 5:
Hardware II: design considerations: power, heat, processing
Guest Speaker: Jonathan Viventi, MSEE, UPenn
Weeks 6 & 7:
Communication in locked-in syndrome: A P-300 based communication syndrome.
*Students choose final projects.
Guest speaker: Gerwin Schalk: BCI-2000, Wadsworth Center, Albany Medical College
Homework: P-300 averaging and processing for spelling words.
Weeks 8 & 9:
“Deep brain stimulation: localizing thalamic networks, and brain stimulation algorithms.
Guest speaker: Stephen Wong, M.D. UPenn or Cameron Macyntire, Cleveland Clinic.
Homework: Sorting and clustering multiunit recordings from implantation of deep brain stimulation electrodes.
Weeks 10 & 11:
Antiepileptic devices. Localizing epileptic networks and responsive devices for epilepsy.
Homework: Detecting epileptic seizures.
Week 12 & 13:
Work on final projects.
Guest speaker: TBA (J&J or Boston Scientific), getting a device from idea to market. The regulatory and approval process.
Guest speaker: John Harris, CEO NeuroVista or Al Hershey, TCD Medical: Start-up companies and entrepreneurship in BCI.
Hands on and help sessions.
Week 14:
Discussion of course. Last lecture on new developments in BCI since beginning of the course.
*Hand in projects.
Projects:
These will build on the homework assignments. Students will choose one of three projects using unknown data sets from the archive. All projects will be graded 1/3 on design, 1/3 on implementation and code, and 1/3 on success at the required task. Students will be given a training set. They will upload code to be sure that it can run on the test set. Test set run will be performed after uploading final code so that final results are fixed and stored.
The demands from the graduate students will be commensurate with their more advanced standing. For example, graduate students will need to localize two structures for the Parkinson's data set, or detect and stimulate two different seizure types, a considerably more difficult assignment (see examples below).
1. Designing a novel BCI device to decode evoked potential data to spell words. Students will first have to decode a training set, then turn algorithm in for decoding a message that is individualized to the student.
2. Localizing subthalamic nucleus coordinates based upon analyzing multi-unit activity. Students will have to give entry and exit coordinates for subthalamic nucleus (STN) based upon streams of real data. Students will be given a training data set for their algorithm.
3. Detecting and triggering stimulation to pre-empt clinical seizures. Students will have to detect a particular seizure pattern then choose a type of stimulation output to suppress seizures. Students will be given a training set. Test set will consist of seizures and artifacts. Successful suppression will require detecting below a specific latency and triggering the right form of stimulation.
Person Preparing Description and Date:
Brian Litt
September 2008
|
