Advanced Cloud Framework for Lung Cancer Diagnosis and Severity Evaluation in Deep Learning

The most recent statistics from the WHO underscore the worldwide impact of cancer, a dangerous ailment claiming 10 million lives annually. In 2020, lung cancer emerged as a predominant contributor to this toll, responsible for nearly 2.21 million deaths. Recognizing the escalating threat of lung cancer, early detection assumes paramount significance for timely intervention and patient survival. This study leverages histopathological images, derived from the microscopic analysis of biopsies, to categorize distinct types of lung cancer. Firstly, the primary focus is identifying Histological Growth Patterns: Acinar, papillary, Solid, and Lepidic patterns using proposed Overlaid Image Analysis, then assigning severity level indexing based on the growth patterns and predicting severity scores through Advanced Signal Processing Techniques. Finally, a severity-based classification performed using transfer learning in Deep Learning and employed EfficientNetB7 to classify the cancer categories: Adenocarcinoma, benign, and Squamous cell carcinoma. Performance metrics such as classification Accuracy, F1-Score, Precision, and Recall are meticulously evaluated. The proposed methodology significantly enhances classification accuracy, reaching 98% to 99.8%, surpassing existing benchmarks. Additionally, deployed this deep learning model into a cloud-based web application to predict the cancer type and its severity. These outcomes give clinicians the means to discern between benign and malignant classifications, facilitating tailored treatment initiation.