MediRay AI: A CNN-LSTM-Based Multimodal Deep Learning Framework for Automated Multi-Cancer Detection from Medical Images

MediRay AI is a research prototype that analyses medical images from five cancer types: lung, glioblastoma, breast, colorectal, and gastric. It uses a CNN-LSTM classifier and reference-based image matching to evaluate uploaded images based on the scan type chosen. On 5000 test images, the CNN-LSTM achieved 92.4% accuracy and a macro-F1 score of 0.908, compared to a baseline CNN's 67.6% accuracy and 0.623 macro-F1 score. These preliminary findings are archive-reported and have not been independently replicated; the dataset split also requires confirmation. The current bundled dataset includes 10,000 synthetic or augmented images, with plans to expand to 50,000 in the future. When an image fails to meet the matching conditions, the system may return an unclassified result. The project also discusses drug-response analysis using gene expression and drug data, though the runnable package currently employs an illustrative scoring method rather than a training regression model. MediRay AI is designed for research and demonstration. Before making claims about clinical performance, validated data, reproducible evaluation, and independent testing are required