Towards Precise Lung Carcinoma Classification: Federated Learning Enhanced CNNs

Sophisticated diagnostic techniques are required for effective and early identification of lung cancer, a worldwide health problem. This study explores the creative use of federated learning and Convolutional Neural Networks (CNN) to tackle the problem. The research comprises data from five different customers and spans five various classes of lung cancer disorders, maximising the adaptability and thoroughness of the learning model. Each client’s data revealed distinct traits throughout the thorough diagnostic examinations. Precision, recall, F1 score, support, and accuracy metrics varied among client datasets, reaching highs of 96.10% precision for Class-5 in the same client and lows of 92.32% precision for Class-l in Client-l, demonstrating the model’s versatility and effectiveness across a range of datasets. To saw a significant improvement in diagnostic performance when local models were combined with a global model using federated averaging. Client-5 excelled with a precision of 93.94% and an accuracy of 0.98, as opposed to Client-l, which had an aggregate precision of 92.14% and an accuracy of 0.97. The consistency of the model was enhanced, and its variance was decreased across client datasets using the federated averaging technique. Federated averaging metrics result analysis: A three-pronged examination that included macro, weighted, and micro averages highlighted the model’s potential even more. With a macro average of 93.S0%, a weighted average of 93.80%, and a micro average of 93.79%, Client-5 had the highest ratings overall. This study clarifies how federated learning may be used with CNNs to improve lung cancer diagnosis. Combining these cutting-edge technologies heralds a potential route for early, precise, and successful lung cancer diagnosis, opening the door for prompt and efficient medical therapies.