Recent Advances in Hybrid Deep Learning for Multi-Class Lung Disease Diagnosis from Radiological Images

Abstract

Lung diseases such as pneumonia, tuberculosis, COVID-19, lung cancer, and chronic obstructive pulmonary disease (COPD) remain among the leading causes of morbidity and mortality worldwide. Effective clinical decision-making depends on early and precise diagnosis using radiological imaging modalities, such as computed tomography (CT) and chest X-rays. Although deep learning (DL) techniques have shown impressive results in automated lung disease detection in recent years, traditional deep models frequently have high computational complexity, large parameter sizes, and limited deployment feasibility in healthcare environments with limited resources. By mixing several learning paradigms and maximizing model efficiency, hybrid and lightweight deep learning systems have emerged as viable answers to these problems. Recent hybrid lightweight deep learning models created for multi-class lung disease diagnosis from radiological images are thoroughly examined in this article. The study systematically examines existing literature by categorizing approaches based on feature extraction strategies, hybrid model architectures, backbone networks, optimization techniques, and classification mechanisms. Performance metrics, datasets, preprocessing techniques, and validation strategies are critically discussed to highlight comparative strengths and limitations of current methods. This research also highlights how lightweight designs can improve computing efficiency without sacrificing diagnostic accuracy, which makes them appropriate for edge-based and real-time medical applications. The study concludes by outlining important research issues, such as clinical integration, interpretability, generalization across imaging modalities, and dataset imbalance. In order to construct robust, scalable, and clinically reliable hybrid lightweight deep learning frameworks for multi-class lung disease diagnosis, future research areas are described. The purpose of this review is to provide scholars and practitioners involved in medical image analysis and intelligent healthcare systems with a useful resource.