Authors: Kharat Yash Amar, Prof. M. C. Kolse, Prof. S. K. Said

Abstract: The growth of digital healthcare has produced large volumes of medical images and clinical reports, and reviewing them manually demands considerable time from radiologists, physicians and records staff. Artificial Intelligence (AI) can support these professionals by analysing images to highlight possible abnormalities and by extracting meaningful information from scanned and typed reports. This paper reviews an integrated AI approach to medical image and report detection. On the image side, it surveys deep convolutional networks for classification, segmentation networks for outlining structures, and visual explanation techniques that show where a model is looking. On the report side, it reviews Optical Character Recognition (OCR) for converting printed and scanned reports into text, layout-aware document models for extracting fields from forms, and biomedical natural language processing for recognizing findings, diagnoses, medications and negations. The link between the two, in which report text is used to label images and image findings are summarized in text, is also examined. A layered architecture covering input acquisition, preprocessing, parallel image and text pipelines, fusion and a clinician-review interface is proposed, together with research gaps, objectives and a comparison of techniques. The review concludes that combining image analysis with report understanding can reduce repetitive work and organize patient information, provided that systems are validated and qualified professionals verify every output.

DOI: https://doi.org/10.5281/zenodo.24218429