Automated Diagnosis of Aortitis Using Machine Learning
CSEF · 2019 Computational Systems & Analysis (Senior Division Only)
Overview
Objectives Aortitis is a rare heart disease characterized by inflammation in the aorta. My goal is to create a reliable tool in diagnosing aortitis by developing an accurate machine learning algorithm to analyze CT scans of the heart. Specifically, the algorithm analyzes the aorta and determine whether it is inflamed. Aortitis frequently goes undiagnosed due to its rarity, so my program will assist doctors in diagnosing it early. Methods To store and compare the CT scans, the python program uses HOG (histogram of oriented gradients) descriptors. They store a histogram of gradients (consisting of x and y derivatives which have direction and magnitude), which is more efficient than storing the entire image because the useful data consists of abrupt changes in the derivatives. In the case of aortitis, likewise, the key difference between an inflamed aorta and a normal aorta would be an increased area with a lesser change in the derivative (due to the thickened aortic wall). For classification, Linear-SVC, an algorithm which establishes a hyperplane between clusters of data, was used. LinearSVC uses the parameters which the HOG descriptor provides to train the program and draw the hyperplane, effectively classifying each image as either having or not having an inflamed aorta. This unique combination of Linear-SVC and HOG descriptors is more accurate than generic deep learning models as it is customized and specific to aortitis. Results With an overall accuracy rate of 94% and a type II (false negative) error rate of only 1.4%, the algorithm proves to be effective. Using the program along with analyzing ESR and CRP levels (two tests in which the levels will be abnormally high in an aortitis patient) makes diagnosing aortitis reliable and less error-prone for doctors. Studies show ESR and CRP to be roughly 96% accurate, so just with my algorithm and these levels alone, Doctors can diagnose aortitis with a 99.8% accuracy. Conclusions I developed a unique and innovative machine learning algorithm which uses HOG descriptors to extract features and Linear-SVC to classify CT scans of the heart to determine whether the aorta is inflamed. In short, I constructed a computer vision program to accurately diagnose patients with aortitis, a rare heart disease. My project expands our horizons in the field of diagnosing rare diseases as well as developing unique machine learning models.
Summary statement
I developed a unique and novel machine learning algorithm to diagnose aortitis, a rare heart disease.
Help received
My AP statistics teacher helped me with the 'Data Analysis' section of my project.
Competition history
- CSEF 2019
Resources
Related projects
CSEF · 2019
Developing a Novel, Accurate, and Rapid Machine Learning Based Skin Disease Diagnosis Algorithm and Mobile Application
ISEF · 2015
It's a Matter of Life and Breath: A Novel Computer Diagnosis Tool for Early Detection of Lung Cancer SPNs towards Big Data Analysis
JSHS · 2024
FRISTS: A Novel AI Framework for Interpretable Heart Failure Prediction Through AI Feature Selection and Time Series Recurrent Neural Networks
CSEF · 2019
Machine Learning Ensemble Model for Improved Personalized Lung Cancer Risk Assessment and Malignant Nodule Detection
CSEF · 2018
A Machine Learning Based Approach to Decrease the Lung Cancer Malignancy Detection Threshold
ISEF · 2019
A Novel Approach to the Diagnosis of Heart Disease Using Machine Learning and Deep Neural Networks
JSHS · 2024
A Holistic Multi-Modal GenAI Healthcare System: Early Detection and Predictive Treatment Monitoring Using Cough Audio and Chest X-rays for Tuberculosis
ISEF · 2021
A Low-Cost Computer-Aided Lung Auscultation Apparatus and Automated Diagnosis of Respiratory Illnesses
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects