How It Works

From uploaded X-ray to explainable model output

MedVision combines study validation, specialist models, validation-derived thresholds, Grad-CAM, and structured reporting.

Step 1

Upload a radiograph

Choose a supported PNG or JPEG chest or musculoskeletal X-ray.

Step 2

Validate the study type

The input is checked before it reaches the specialist analysis model.

Step 3

Run model inference

The selected deep-learning model generates structured image-level outputs.

Step 4

Generate Grad-CAM

A heatmap highlights regions that influenced the model output. It is not lesion segmentation.

Step 5

Review the report

Results, metrics, limitations, and downloadable reports are presented in one workflow.

Important limitations

Model outputs may not generalize across scanners, institutions, patient populations, acquisition settings, or unsupported study types. MedVision is an educational research project, not clinical software.