Machine Learning Based Algorithm Development for Cardiovascular Biomechanics

Project description

The aim of this project is to similarly create a system that can use the micro vibrations produced by fluid movement, and be able to measure the volume and flow of the fluids. This would have wide ranging applications, with fields where direct measurements of the volume and flow may not be feasible due to lack of access, as in the case of remote systems like micro irrigation systems or toxic materials/ high temperatures like in the case of measuring fuel efficiency. For instance, presently, engine manufacturers use a tracer gas to estimate the volume of fuel within the ignition chamber. [5] However, a contactless system can provide accurate measurements of flow without the use of such tracers. Another instance where direct measurement is not possible is the human body, where blood flow through the heart, or air movement through the lungs cannot be directly measured easily. The Dozee device currently uses BallistoCardioGram (BCG) signals to capture HR and RR, and in doing so, provides a cost effective and noncontact way for measuring such vitals. Currently, the technology to measure volume in a similar contactless manner remains elusive, and this project aims to fill this gap.

Outcomes

• Literature review along with brainstorming with Medical experts to derive a hypothesis. Clinical significance, usability and practicality of the hypothesis to be considered before development
• Identify DSP techniques and ML algorithms that can be used on vibration data to cluster and classify signals across multiple parameters/markers
• To design a protocol to utilize existing and acquire more training set data with cardiac output clinical data
• To optimize the sensor array position, sampling rate and sensors to be able to acquire the necessary signals
• Modeling of the pulse transit based on multiple sources to identify the scope of screening for cardiovascular disorders – heart failure, hyper/hypotension, etc.
• Create a validated algorithm layer to the base BCG program and demonstrate proof of concept with known ground truth and simulated data sets acquired with the clinical partner institute
• Explore other methods non-obtrusive methods to evaluate cardiac output

Essential capabilities

An ideal student should be interested in computational (bio)mechanics, AI/ML based algorithm development, bio signal processing. The project is mostly going to be on computational cardiovascular engineering and its implication in cardiac healthcare domain.

Knowledge in MATLAB or Python will be required.

Desireable capabilities

Interest in table top hardware building, data acquisition etc would be desirable.

Expected qualifications (Course/Degrees etc.)

Preferably an MTech in Mechanical Engineering or Biomedical Engineering, and any other computational science.