Clinical Trials Lead and Principal Engineer - Analog Devices Inc
(2020-09)
- Design of experiments to validate cloud based algorithms working on physiologic signals, recorded from wearable (chest worn) cardiovascular monitor device
- Design of a clinical rule based Triage algorithm to escalate patients that need immediate attention for cardiac arrest conditions -- Currently being tested out on multi-clinic trials.
- Leading a team of two members and two interns and indirectly leading multiple members of SW and Algorithm team on 2 different clinical projects.
- Handling clinical studies with a budget of more than $4.0 millions. Study Partners @ University of Glasgow, Vanderbilt University, and New York Presbyterian Hospitals. Converted the relation to setup a Center of Excellence with University of Glasgow.
Lead Algorithms Engineer - Philips Connected Sensing Business Unit
(2016-03 - 2020-08)
- Led the development of real-time embedded algorithms in a wearable platform (ARM-4) Computation of following parameters in real-time -- Heart Rate, Respiration Rate, Posture, and Ambulation and Activity Level.
- Computed the measurement errors and logs the real-time information on device and patient contact issues.
- Led the complete design cycle for algorithms – conceptualization,design, development and V&V.
- Led a direct team of 4 algorithm/SW engineers and 3 interns and indirect team of 5 engineers at Eindhoven R&D facility of Philips.
- Resulted in two successful 510k submissions and two IP application.
Algorithm Consultant - Avicena, CalTech Startup
(2015-11 - 2016-06)
Developed user-interface/ patient friendly algorithm for waveform quality measurement in a novel pulse and blood pressure measurement systems for short term and long term monitoring technology.
Research Specialist - Department of BME, UCI
(2015-08 - 2016-03)
- Designed multi-channel signal processing solutions to aid the analysis of multi-electrode array recordings from cultures of neuron cells from animal brains.
- Studies include artifact rejections, spike detection and estimate of communication between different chambers of hippocampus cells harvested from rats brain.
- Work resulted in two publications.
Consultant Algorithm Engineer - Axonics Modulation Systems, USA
(2014-12 - 2015-12)
- Designed filter chain and signal processing component for an EMG system as part of clinician neuromodulation equipment.
- EMG used to measure the placement accuracy of implantable device to treat urinary incontinence.
Senior Algorithm Engineer - Infobionic Corporation, USA
(2014-06 - 2015-06)
- Developed algorithms to estimate heart rate variability measures and to analyze the ECG and respiration signals for movement artifacts. Creating EC-57 reports and V&V framework for the cloud-based algorithm performance.
- Contributed to 510k approval from US and EC certification from European regulatory agencies.
- Modifying the ventricular beat feature set to create a more effective feature set and to improve beat detection efficiency. Set of 30 features over a huge template set in the cloud is used to classify ventricular beats, normal beats, and other beats.
- Used three-axis accelerometers (ADS and STMicro) recordings to compute the level of physical activity for patients and to classify them to multiple levels.
Senior Principal Algorithm Engineer - Cardiac Science Corporation (now part of ZOLL), USA
(2008-06 - 2014-06)
Development of real-time signal processing algorithms to measure and remove motion artifacts present in physiological signals, during cardio-pulmonary resuscitation (CPR).
Piecewise Stitching
Adaptive (PSA) algorithm is a real-time, deconvolution algorithm that separates bio-signals from artifacts. Trans-thoracic impedance based CPR detection has been a by-product of this effort. Also, ECG based CPR detection has been added. A PC based demo is followed by animal trials at UCLA, Emergency Cardiology Lab.
- Machine Learning 1: Energy efficient computation for wearable AED and Monitors - Certain features are utilized in multi-channel physiological systems to adaptively increase or decrease the signal length for computation. Preliminary models indicate almost 80% gain over continuous monitoring devices in terms of computational need and battery use. General model optimal for any multi-channel, multi-time series analysis.
- Machine Learning 2: Field Data Analysis and Test Data Extraction towards Pre-Market Approval process - Signal preparation and evaluation of field performance for Rhythm classifier for over 1200 rescue data and 4000 simulation data available with the company. Signal preparation involves inverse filtering, digital to analog conversion, signal calibration and use of automated database test fixtures.
- Completed Project 1: Verification and Validation protocol for Arrhythmia Detection Algorithm - Completed the development of a Tool -- RhythmX Performance Analysis (RxPA) software to automatically compare and score the performance from 74 hours of clinical data set. Complete automation of this protocol saves 90% time for the company, with minimal manual tester requirements. Added multiple synthetic signals to complete algorithm test requirements. This is being utilized by all CSC AED product line
- Completed Project 2: Verification and validation of filter sections for ECG and CPR Assist Device (CAD) modules for digital filter effects - Performed an analysis of finite word length impact on stability and limit cycle characteristics to understand the software digital filters. Analysis resolved the issues with multiple second-order sections (SoS) in prototype AED and CAD devices.
- Completed Project 3: Prompt Engine Design for CPR Assist Device (CAD) - Incorporated in next generation product, this module computes the depth of chest compressions and prompts the first-responders to adjust their compression rate and depth. An innovative weighting scheme is used to account for pauses and rapid changes in compression. A C# interface was designed to create a real time simulation of CAD interface with AED. Depth measurement was achieved using accelerometer based CAD.
- Completed Project 4: Rhythm analysis algorithm decision rule(s) and its impact on offline model - A device equivalent offline rhythm analysis model was created to compute the impact of changes in rules for "Shock/No Shock" decision. This led to modifications in a particular variant of Cardiac Science AEDs.
Professor and Head of the Department, Advanced Embedded Systems Design Program - International Institute of Information Technology, Pune, India
(2004-06 - 2008-06)
- Personal focus was on creating and embedding signal and image processing algorithms on integer point machines, under platforms like DSPs, ARM Core and FPGAs.
- Managed a team of 10 faculty, 8 research associates and graduated 110 Master level students (practicum and research). Managed two academic research labs, one on embedded controllers and another on embedded computing.
- Editor and Coordinator, NCECCube -- National Conference of Embedded Computing, Control and Communications, 2007.
- Patni Seminar Series on "Physiology of the Heart" for Engineers, 2005-2007.
- Imaging and Vision Seminar Series, Sponsored by Patni. (2007)
- Post-graduate and PhD advisor/ examiner for IIT, Chennai, Anna University, Devi Ahilya University, Indore and University of Pune.
- National seminar series in 2007/2008 in Embedded Computing.
- More than 20 research publications in theoretical and embedded signal and image processing.
Algorithm Engineer/ Consultant - Cardiac Science Corporation, Irvine, USA
(2000-10 - 2003-10)
- Part of the development team in company's first bedside AED.
- Analyzed the possibility of bedside AED expanding its role as monitor device and as a predictor of sudden cardiac death.
- Developed an algorithm for detection of microvolt T wave alternans in resting and stress ECG.
