A Hybrid Deep Learning and Ensemble Approach for Nodular Goiter Diagnosis using YOLO and AdaBoost
Natrayan L, Kiranmai Doppalapudi, Guduri Padma Rao, S. Kaliappan, Ramya Maranan, Siva Kumar Pathuri
Abstract
In the past couple of years, a sharp increase in the rates of goiter and thyroid cases all over the world has been recorded and cases have been found in all increasingly grand age groups. Amidst the fact that the thyroid gland plays a vital role in controlling important metabolic activities, it is vital to diagnose abnormalities efficiently and at the initial stages. Nonetheless, a number of current diagnosis algorithms either specialized in a small number of two thyroid diseases or do not take the class imbalance of datasets into consideration. This paper presents a dependable and modifiable system by incorporation of AdaBoost classifier incorporated in YOLO (You Only Look Once) system that detects objects at high speed. The model proposed can correctly identify ten comports of goiters and thyroid malfunctions in one framework. It has an impressive 96 percent accuracy ranking higher than other conventional classifiers such as Naive Bayes (88 percent) and Logistic Regression (91 percent). Due to the successful address of the problem of its data imbalance and the complicated medical trends, the YOLO+AdaBoost system brings high diagnostic accuracy. This will be effective in healthcare across the world, as it will encourage early diagnosis and better treatment outcomes on goiter and thyroid diseases.