OncoVision-AI: Unified Multi-Organ Tumour Screening using CNN–Transformer Models

Early and reliable tumour screening often requires separate diagnostic tools for different organs, resulting in fragmented workflows and inconsistent user experience. Also, this work introduces OncoVision-AI, a web-based platform designed to perform multi-organ neoplasm screening using deep learning. The system brings together individually trained classifiers for lung MRI, brain MRI, osseous tissue X-ray, and eye fundus images, each optimized to its respective modality while sharing a common deployment layer. Evaluation crossways four datasets show strong performance for brain, osseous tissue, and lung tumour detection, while the eye Cancer task highlights the effect of severe class imbalance - a finding that guided our error analytic thinking. All models are integrated into a single interface that allows for organ selection, image upload, confidence scoring, and visualization of outcomes. The results prove that modular, organ-specific training combined with a unite delivery environment can simplify deployment and support practical, multi-organ screening scenarios. Future enhancements include calibration techniques, dataset rebalancing, and interpretability features to improve reliability across all endpoints.