By Hana Suzuki, Medical Technology Correspondent
Dhaka — October 8, 2025

A new artificial intelligence system developed by Bangladeshi Computer Science student Samin Yasar Al-Sami aims to revolutionize medical imaging in rural hospitals. The platform can analyze X‑ray and MRI scans, highlight anomalies, and prioritize urgent cases, assisting clinicians in areas where specialist radiologists are scarce. Discussions are underway with Shibukawa Rural Hospital in Gunma Prefecture, Japan, for a pilot program to evaluate the system’s clinical impact.

The AI system is designed as a clinician-assist tool, providing guidance without replacing medical professionals. It combines a classification ensemble, using EfficientNet and Swin Transformer backbones, with a modified U‑Net segmentation network. This architecture allows the platform to generate case-level probability scores and pixel-level heatmaps that localize anomalies, giving clinicians both a prioritized case list and visual guidance for rapid decision-making.

Training involved public datasets, including NIH ChestX-ray and MIMIC-CXR, along with a curated collection of anonymized X‑ray and MRI images. Internal validation produced high accuracy across multiple conditions, with rapid inference times enabling near real-time analysis. On a standard CPU, results are delivered in under ten seconds per image, while GPU inference takes less than two seconds. The system supports standard DICOM formats and integrates seamlessly with hospital PACS and electronic health record workflows.

Rural hospitals often face delays in diagnosis due to limited radiology staff. By pre-screening scans and flagging urgent cases, the AI platform helps clinicians focus attention on high-priority patients. The interface displays original scans alongside heatmaps and a summary of probable findings, providing an intuitive tool for workflow integration.

Shibukawa Rural Hospital officials described the system as promising, emphasizing that human oversight and patient privacy remain essential. “Tools that responsibly highlight urgent cases and support clinician decision-making are valuable for our hospital,” said a hospital spokesperson.

The proposed pilot includes a retrospective analysis of historical scans to benchmark AI predictions against radiologist assessments, followed by a three-month monitored clinical trial where the AI functions strictly as a decision-support system. Clinicians will provide feedback to refine the platform, and metrics such as time-to-triage, concordance rates, and operational performance will be recorded.

Experts say AI solutions in healthcare hold significant promise if validated locally. Dr. Keita Nakamura, a radiologist not involved with the pilot, noted, “Models trained on large datasets must be adapted to the hospital’s imaging protocols and patient population. A careful staged evaluation is the right approach.”

The developer emphasizes that the AI platform is intended to enhance human decision-making, not replace clinicians. Any future deployment will follow ethical standards and Japanese regulatory requirements. If successful, the system could serve as a model for rural hospitals throughout Japan and other regions facing similar healthcare challenges.