MFANEDH: Multimodal Fusion of MRI, PET, CT, and Ultrasound Using Advanced Neural Architectures for Enhanced Early Detection of Heart Cancer

Heart cancer poses unique diagnostic challenges due to the complexities of the cardiac environment and the subtle nature of tumor manifestations. Timely detection is crucial given the increasing incidence of heart cancer cases. Current imaging technologies, while providing significant advancements, often operate independently, limiting the utilization of diagnostic information. Integrating sense modality such as MRI, PET scans, CT imaging, and ultrasound can overcome these limitations. In response to the fragmented diagnostic viewpoints observed in clinical practice, a multi-modal fusion framework was developed. This framework utilizes deep learning models tailored to each modality's strengths: 3D CNNs for MRI, U-Net combined with ResNet for PET scans, DenseNet for CT imaging, and BiLSTM for ultrasound. By amalgamating these models, the framework effectively leverages the unique features of each modality, resulting in substantial improvements across key diagnostic metrics. The context achieved a 3.5% increase in precision, 2.9% in accuracy, 2.5% in recall, 4.5% in AUC, and 4.9% in specificity. Additionally, it led to an 8.3% reduction in diagnostic delay, thereby enhancing diagnostic heftiness and ensuring timely detection. This research represents a paradigm shift in heart cancer diagnostics, laying the groundwork for integrated imaging platforms and refining patient outcomes and modified treatment strategies.