Thyroid Disease Prediction using Quantum Support Vector Machine Classifier

Abstract

Thyroid illness is a significant health issue necessitating precise diagnostic methods to enhance treatment efficacy. Conventional approaches often encounter difficulties stemming from intricate data properties, such as class imbalance and noisy features. A Quantum Support Vector Machine (QSVM) classifier was used to overcome these constraints in predicting thyroid illness, capitalizing on the computing benefits of quantum principles for managing multidimensional feature spaces. Two datasets were employed: the UCI Machine Learning Repository thyroid dataset, consisting of about 2,800 training instances and 972 testing instances with 29 characteristics, and the DDTI: Thyroid Ultrasound images dataset, which includes 480 ultrasound images annotated by clinical specialists. The UCI Machine Learning Repository dataset was used to assess parameter-based classification efficacy, whilst the DDTI images offered imaging-based validation for robustness. The QSVM classifier attained a prediction accuracy of 99.44%, precision of 98.62%, recall of 98.61%, and F1-score of 98.61%, underscoring its efficacy in distinguishing between normal and abnormal thyroid diseases. The findings indicate that quantum-inspired learning frameworks may enhance clinical decision support systems for predicting thyroid illness.

Ähnliche Arbeiten