Deep Learning Models for Medical Image Classification: A Comprehensive Review

Abstract

The field of medical informatics involves the study of integrating imaging and biomedical record data. Medical image data consists of pixels representing different parts of a physical object. Analyzing this data requires expertise in value analysis and disease diagnosis. Medical image classification is crucial for Computer-Aided Diagnosis (CAD) and improving healthcare services. It involves analyzing pixel data to categorize medical images and identify affected areas. This is challenging due to the high dimensionality and complex structures of medical images. Experts are needed to interpret image features and verify classification results. The key goal is to maximize categorization accuracy for precise disease diagnosis. In recent decades, Artificial Intelligence (AI) models, including Machine Learning (ML) and Deep Learning (DL) algorithms, have been developed for medical image classification. Traditional ML relies on hand-engineered features, while DL models automatically extract discriminative features at multiple levels of abstraction. However, challenges such as limited training data, class imbalance, and inter-class similarities hinder the learning of salient visual characteristics. This study provides a thorough examination of the most advanced DL models used for classifying medical images. Furthermore, it evaluates the advantages and constraints of various models on a range of medical image datasets. The review provides valuable insights on how to optimize medical picture categorization and offers guidance for future improvements in Computer-Aided Diagnosis (CAD) to improve precision healthcare.