Deep Learning-based Staging of Throat Cancer for Enhancing Diagnostic Accuracy Through Multimodal Data Integration

Aim This study strives to develop deep learning models with respect to the staging of throat cancer through CT image analysis linked with medical records. Using 650 CT scans, the current research combines the convolutional neural networks of VGG16, VGG19, and ResNet50 with K-Nearest Neighbors (KNN) for text data analysis to enhance diagnostic accuracy. On the other hand, several such implementations have shown potential in increasing the process of clinical decision-making by integrating image and textual data, thereby providing great potential for deep learning in medical diagnostics. Background This study investigates the performance of deep learning in staging throat cancer by CT images and clinical records. A dataset of 650 CT scans was processed using advanced Convolutional neural networks (CNN) by VGG16, VGG19, and Res-Net50 integrated with KNN for text-based data. The most accurate model was the VGG19+KNN model (98.67% accuracy), which proved that the fusion of multimodal data will result in a diagnostic enhancement in terms of precision for medical imaging and cancer diagnosis. Objective The study determined the effect of deep learning models in VGG16, VGG19, ResNet50, and KNN combined with stage throat cancer by using CT images and clinical records. It is 98.67% by VGG19+KNN, pointing at the great promise for accurately diagnosing cancers. Consequently, fusing image and text data have been included into a single system for decision-level fusion in such a way that significant improvement has been observed with regard to diagnostic precision, thus providing support to the views of other researchers about the potent positive effect that deep learning could bring in the area of medical diagnostics. Methods In the current study, 650 CT scans were analyzed for effective staging of throat cancer by the proposed deep learning method. The use of VGG16, VGG19, and ResNet50 convolutional neural networks helped in extracting features from the CT images. Analysis of the corresponding clinical text data was done using KNN. For improving the diagnostic accuracy, decision-level fusion was performed by integrating the models. The model VGG19+KNN showed the best results of 98.67%. Results This research evaluates deep learning models for the staging of throat cancer using CT images and medical records. A total of six hundred fifty CT images were enrolled, where CNNs VGG16, VGG19, and ResNet50, along with KNN for clinical data, were utilized in this study. The highest accuracy obtained was by the VGG19+KNN model, which is 98.67%, compared to the VGG16+KNN model, which showed 94.5%, followed by ResNet50 with an accuracy of 92.3%. Therefore, the results postulate that the integration of imaging and textual data improves diagnostic accuracy in cancer staging. Conclusion This study showed a good level of performance of deep learning models, especially the VGG19+KNN model, in accurate staging of throat cancer using CT images and clinical records. The highest classification accuracy was achieved by the VGG19+KNN model, markedly improving the diagnostic precision. It strongly suggests that multimodal data integration—that is, imaging with text analysis would lead to better clinical decision-making in cancer diagnostics. This explains the transformational power of deep learning in medical diagnostics.