Wednesday, 18 December 2013

Initial project proposal

I recently joined innovation center at LVPEI as a Research Associate to work on electrophysiological testing of the eye. So I'm share research proposal as given under. Please review and share your opinion.

General purpose, high-performance system for clinical electrophysiological testing based on modern computational techniques featuring low cost, extensible framework and enhanced user experience
 
Abstract — Continuous improvements in noise performance of analog frontends resulting from advancements in IC fabrication processes and growing availability of powerful computers in form of smartphones, tablets and personal computers allowing use of computational techniques such as digital signal processing and machine learning suggests a need to reconsider the approach for clinical electrophysiological testing. Performance enhancements in these systems can enhance its user experience by reducing testing time and invasiveness for a test subject. There is a substantial benefit in designing a general purpose system as most systems for electrophysiological testing have significant similarities in their operating principles, differing only in biopotential signal levels, presentation of test results, type and placement of electrodes, user interface and test specific stimulus if applicable. A general purpose system for electrophysiological testing is therefore being proposed to offer substantial functional benefits and flexibility over existing commercial systems. It will be first demonstrated for visual diagnostics in the best interest of LVP-MITRA innovation center at L V Prasad Eye Institute, Hyderabad where this research is being carried out.

A quick self-explanatory block diagram


The Electrodes, DAS/GPIO and Stimulus blocks constitute the hardware. Test definition, Digital filters, Supervised learning and UI blocks are implemented in software. Electrodes are placed suitably on a patient or test subject and a stimulus if needed as in case of ERG and VEP is set up. The clinician selects an appropriate test from the UI for which a test definition is available.

Test definition:
Test definitions are test specific algorithms that use all other subsystems as required for a specific test. This is the heart of extensibility of the design. It also includes how the result is being presented and what controls are made available in the UI when a particular test is selected.

Data acquisition and General Purpose IO (DAS/GPIO)
DAS/GPIO is a hardware component that contains the analog frontend with a dynamic range large enough (>106dB) for capturing all biopotential signals in 5mV to 0.1uV range, avoiding clipping and recovery delay. 'Most' filters will be implemented in software as digital filters have better roll-off and phase characteristics and do not add to the cost. To compensate for and measure DC offset at the inputs of the biopotential amplifier and to preserve its dynamic range, the amplifier design has auto-zeroing technique. The GPIO block is used allow test logic to exchange data with stimulus generator or interactive controls if used. To comply with safety standards it also uses digital isolators to prevent the test subject from becoming a part of the mains circuit. 

Supervised learning:
Supervised learning is a machine learning algorithm that can be used to: 
  • Identify useful features in waveforms
  • Identify and ignore invalid waveforms when taking multiple trials for applying signal averaging. This can considerably reducing testing time.
  • To provide quick screening tests based on how a normal signal looks like.
Machine learning relies on large amounts of data gained from experience, so there is a need to integrate training routines into the UI. The UI can take short surveys during each test to effortlessly train the system with normal use. Features based on machine learning will stay inactive unless the system has been sufficiently trained.