Authors:

Priyadarshini Subramaniam

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Diffusion coefficient measurement has the potential to become important in many of today's medical diagnostic devices. With the advent of at-home, accessible testing, it becomes necessary to optimize existing particle- based techniques of diffusion coefficient measurement, specifically particle diffusometry (PD) methods. PD utilizes pictures of particles taken at varying time points to measure the diffusion coefficient, with the potential for disease diagnosis. Yet, the performance of existing PD methods degrades drastically in the presence of vibrations, a common occurrence during in-field applications. Although physical vibrational suppression methods (such as an optical table) and software-based vibration suppression methods (using fiduciary marks and subsequent frame-by-frame shifting) is available, the high-cost barrier of optical tables and absence of available fiduciary marks in micro-PD systems makes these methods ineffective. This work attempts to use single particle tracking and deep learning algorithms to suppress in-field vibration of PD tools. This technique utilizes various vibrations to create image datasets of particles through time and uses a convolutional neural network to predict the underlying diffusion coefficient. The results are expected to show how to account for long-term consistent vibrational motion of a PD device using existing and new deep learning models. It is expected that deep learning models can be trained to account for the vibrational motion of the particles. Based on the findings of this study, long-term consistent vibrational noise is expected to be filtered out using deep learning models.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

Priyadarshini Subramaniam