Performance analysis of deep learning architectures for chest disease and lymphoma classification

Integrating deep learning and federated learning models, the study investigates the detection of cancer and chest diseases.  The researchers used a combination of 112,120 chest X-ray photographs annotated by the NIH and 5,400 lymphoma biopsy images from Kaggle to identify three distinct forms of lymphomas.  The method's first aims are to improve images, normalize data, and identify features.  Afterwards, it evaluates features that rely on contours, such as aspect ratio solidity and intensity fluctuations.  The study included seven industry-standard models and federated learning techniques.  The InceptionV3, MobileNetV2, DenseNet161, ResNet50, VGG-19, and VGG-16 models are included here.  To measure performance, we employ F1-score, recall, accuracy, and computational efficiency.  Although InceptionV3 dominated in terms of loss and root-mean-squared error, DenseNet161 had the best accuracy among deep learning models used to detect chest illnesses at 88.01%.  for compared to other models, VGG-19 had a superior accuracy rate of 97.5% for classifying lymphomas.  The newly integrated models that succeeded were InceptionV3 and VGG-19, which had 95.8% accuracy in diagnosing lymphoma and 97.7% accuracy in detecting chest illnesses.  With the use of deep and federated learning algorithms to medical imagery, automated illness detection got increasingly accurate.