Go to Polygence Scholars page
Anthony Novokshanov's cover illustration
Polygence Scholar2024
Anthony Novokshanov's profile

Anthony Novokshanov

Class of 2025Georgia

About

Hello! My name is Anthony Noovkshanov and I decided to write my research paper on using explainable AI to locate pneumonia. I chose this topic due to my interest in how artificial intelligene and computer science can be used to improve the lives of everyday people.

Projects

  • "Using Explainable Artificial Intelligence to Locate Pneumonia" with mentor Vidya (Jan. 17, 2024)

Anthony's Symposium Presentation

Project Portfolio

Using Explainable Artificial Intelligence to Locate Pneumonia

Started Sept. 19, 2023

Abstract or project description

Artificial intelligence (AI) has already become a vital resource in numerous industries; however, it is often challenging to understand how AI reaches its results. This lack of transparency, combined with potential biases within the machine learning model, prevents professionals in critical fields like healthcare from relying on deep learning models for diagnostic purposes, hindering the widespread use of AI in healthcare. This paper investigates the application of XAI in the medical field, focusing on the detection of pneumonia through the analysis of lung X-ray scans. In this study, we developed an XAI tool, utilizing a Convolutional Neural Network (CNN) constructed with PyTorch and trained on the Pneumonia MNIST dataset. Our model achieves an accuracy of nearly 91% on 28x28 pixel images and highlights the top pixels considered most important by the deep learning model in its decision-making process. The primary aim of this project is to present a proof-of-concept tool for the integration of XAI into healthcare diagnostics, with the goal of assisting medical professionals in making informed decisions and ultimately saving lives. By demonstrating the feasibility and effectiveness of XAI in pneumonia detection, we lay the groundwork for future advancements in healthcare AI, emphasizing the importance of transparency and reliability in AI models.