<p>Lung illnesses, and lung cancer in particular, continue to be the world's leading causes of mortality and morbidity because of their intricate pathophysiology and the difficulty in obtaining an accurate early diagnosis. The accuracy, specificity, and current computation of traditional diagnosis are hindered by inherent limits. The Internet of Medical Things (IoMT) and recent developments in deep learning have created new opportunities to tackle these difficulties, however the current methods suffer from issues like low diagnostic performance, inadequate feature representation, and lack of interpretability. This research presents a novel approach for the early diagnosis of lung cancer from IOMT CT scans: Simplicial Cascaded Deep Capsule Neural Network with Chef Leader-Based Optimization (SCDCNN-CLBO). Graph Enhanced Fuzzy Clustering (GEFC) is used for precise segmentation after the Adaptive Morphological Wavelet Perona–Malik Filter (AMWPMF) is used as pre-processing to increase image quality. The classification process employs the SCDCNN model that integrates Simplicial Attention Networks (SAN) and Cascaded Deep Capsule Neural Networks (CDCNN) for discriminative feature learning and hierarchical representation of spatial relations. Optimization is carried out through the Chef Leader-Based Optimization (CLBO) to enhance classification performance even more. Experimental results prove that the approach proposed here attains 99.9% accuracy in diagnostics and 99.8% precision and, therefore, it is a very efficient and dependable solution for real-time early lung cancer detection in IOMT-based health environments.</p>