The Bridge - Issue 2, 2023 - 23

Graduate Research SP
Foram Sanghavi
Epsilon Delta
Tufts University, Ph.D. Student in Electrical Engineering
RESEARCH TOPIC
Artificial Intelligence based Disease Recognition Systems
Leveraging Artificial Intelligence concepts for effective disease prediction has always been
the focal area of my research. The main aim of developing and deploying the AI model
is to help radiologists/pathologists make quicker and more accurate disease predictions
by overcoming human challenges that can affect diagnostic decisions. These challenges
include fatigue, prediction and tissue staining variability, and visual or feature similarity
between diseases. During my Ph.D., with my advisor, Dr. Karen Panetta, 2019 IEEE-HKN
President, we are striving to develop tools for diagnosing diseases such as COVID-19,
cervical, prostate, and breast cancer. My work centers on understanding the critical disease
attributes such as textures, cell distribution, and morphologies essential for disease
detection to develop different disease-diagnostic systems. This was achieved by developing
different texture and morphology feature descriptors to extract complex critical information
for better disease prediction. Our developed AI tools have demonstrated 98.44% sensitivity
for distinguishing COVID-19 x-rays from viral pneumonia X-rays [1]. This was challenging
for the radiologists to perform with the naked eye, as the symptoms and visual textures
look very similar for both diseases in the images. Similarly, our system has been shown to
deliver > 95% accuracy in predicting cervical and breast cancer tissue images.
[1] Panetta, Karen, et al. " Automated detection of COVID-19 cases on radiographs using
shape-dependent Fibonacci-p patterns. " IEEE Journal of Biomedical and Health Informatics
25.6 (2021): 1852-1863.
LEARN MORE
https://www.karenpanetta.com/lab-members
TLIGHT
The figure illustrates the different feature descriptors developed
for different disease types. (Top-left) A Covid-19 chest x-ray
(Top-middle) is a Shape-Dependent Fibonacci image, a texture
feature descriptor image that extracts necessary disease
information from lungs, and (Top-right) is a histogram of this
image used for diagnosis. (Bottom left) is a Prostate Cancer
biopsy image, (Bottom -middle) is cell segmented image
from which texture, morphological and spatial information is
extracted, and (Bottom-right) is a spatial graph constructed
on the cell-segmented image from which spatial distribution
features as obtained and used for classification.
CONTACT
https://www.linkedin.com/in/foram-sanghavi-375507b0/
HKN.ORG
23

https://www.karenpanetta.com/lab-members https://www.linkedin.com/in/foram-sanghavi-375507b0/ https://grad.tufts.edu/engineering-masters/ https://hkn.ieee.org/

The Bridge - Issue 2, 2023

Table of Contents for the Digital Edition of The Bridge - Issue 2, 2023