Deep Learning-Based Diagnosis System for Liver Cancer Through Multimodal Determination for Early Treatment

The present research provides an analysis of the development and implementation of a deep learning-based diagnosis system for liver cancer. The provided system was developed using a dataset encompassing 2300 sensor readings, including imaging, clinical, and genomic data. A range of deep learning features was implemented to improve the early detection of liver cancer such as CNN, VGG 16, VGG 19, and Inception V3. The data were divided into two categories: one is for training, and another used for testing. The results of the data analysis have shown that VGG 19 provides the highest diagnostic accuracy of 98.76<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, with VGG 16, Inception V3 and CNN displaying the levels of 94.5%, 93.4%, and 90.25%, respectively. The Model performance analysis was also augmented at the application of precision, recall, and F1 score indicators, as well as using AUC-ROC values which highlighted VGG 19 as the most effective algorithm. Another method of evaluation was the application of confusion matrices for plotting the results based on the number of true positive, true negative, false positive, and false negative detections, where all the samples from the dataset were added to each model calculation. The obtained results have shown the importance of multimodal data for the increase of diagnosis accuracy and revealed the promising potential of advanced computation techniques in medical analytics. In such a way, it can be stated that the developed deep learning models contribute to the ability of healthcare specialists in the definition of liver cancer at early stages, which allows timely interventions and facilitates the development of a range of personalized treatment protocols. The application of such algorithms has the potential to reshape liver cancer diagnostics, where major tasks are associated with further improvement and validation of the developed algorithms based on clinical case studies.