Multi-Modal Attention-Based Temporal Graph Convolutional Network (ATGCN) for Real-Time Cancer Progression Prediction Using the TCGA Lung Cancer Dataset

To allow early diagnosis and effective treatment, it is essential to have an accurate prognosis of the progression of cancer. This permits for personalized medical interventions which can lead to enhanced patient outcomes. The currently existing method have many limitations such as difficulties in real time application, high computational complexity and complexity in balancing multi modal data from different sources. A unique multi modal attention based temporal graph Convolutional Network (ATGCN) is purposed for predicting the course of cancer to address this challenge. The attention-based method of the ATGCN method allows it to obtain the temporal evolution of cancer cells when integrating the spatial dependencies among cancer cells with their biological connections. This is performed through the use of Graph Convolutional Network (GCN). Hence there is low requirement of resource intensive recurrent neural network (RNN) which results in method being computationally efficient and suited for therapeutic in real time. Advanced pre trained methods such as EfficientNetv2s AND NASNet are utilized to increase the feature extraction ad multi modal fusion. These methods are used also to capture high level features from genomic and histopathology data simultaneously. Then, the features are integrated at the node level to produce dynamic graphs which indicates the evolution of cancer over periodic time. Method optimization is accomplished by the use of RMSProp which ensures effective training and rapid convergence. The proposed ATGCN method is analyzed using TCGA Lung Cancer Dataset and it displays the enhanced performance in predicting the cancer stages in comparison with existing methods. The proposed method outperforms the existing methos in terms of relevant metrics and enhance the accuracy of prediction of method. This method is suitable for real time environment and gives clinically timely insights to the evaluation of cancer.