Implementation of Artificial Intelligence In A Computer-Aided Diagnosis System For Disease Detection from Medical Images

The Chest X-ray (CXR), a widely utilized and accessible projectional radiograph, serves as the primary data input for the proposed Computer-Aided Diagnosis (CAD) system for pulmonary disease detection. The image itself is generated based on the principle of differential X-ray attenuation, where dense structures like bone appear white (radiopaque) and air-filled spaces like the lungs appear black (radiolucent), providing critical visual features for AI interpretation of conditions such as pneumonia or tuberculosis. In the context of this research, the CXR is acquired and stored in the DICOM standard, ensuring the preservation of both image data and necessary metadata. This initial DICOM file triggers the entire operational workflow: it is processed by the multimodal MedGemma 4B model for automated analysis and report drafting, and subsequently, the final verified diagnosis is transmitted via the HL7 FHIR standard, guaranteeing seamless integration with the national SATUSEHAT platform and transforming the image from a simple diagnostic tool into a fully regulated and synchronized data asset.