How jitheswar/sih26038 works
This is a MATLAB diabetic-retinopathy screening product: a fundus image is assessed for quality, graded, checked against lesion and explanation evidence, then assigned a screening disposition. The README also documents a single-case demonstration app and a separate district-capacity simulation. Relationships below follow those documented capabilities; source excerpts were unavailable, so internal call wiring beyond the documented workflow is not asserted.
Main entry points: runScreeningCase.m Case runner · decisionPolicy.m Screening policy · infer.m Calibrated grading · buildLesionEvidence.m Lesion evidence
Components 17
Screening workflow
- Case runner
src/+app/runScreeningCase.m - Image preprocessing
src/+common/preprocess.m - Image quality
src/+quality/assess.m - Calibrated grading
src/+grade/infer.m - Probability calibration
src/+grade/applyTemperature.m - Screening policy
src/+grade/decisionPolicy.m
Evidence channels
- Grad-CAM evidence
src/+explain/gradcam.m - Learned lesion channel
src/+segment/segmentLesions.m - Classical lesion channel
src/+segment/detect.m - Lesion evidence
src/+explain/buildLesionEvidence.m - Spatial agreement
src/+grade/spatialAgreement.m
Product interfaces
- Screening app
app/ScreeningApp.m
Capacity simulation
- Simulation runner
simulink/runDistrictSimulation.m - District capacity model
simulink/district_model.slx - District configuration
simulink/district_config.json
Other components
- Screening user
- Fundus photo
Connections 20
- Screening user → Screening app: uses
- Fundus photo → Screening app: provides image
- Screening app → Case runner: demonstrates case
- Case runner → Image preprocessing: processes image
- Case runner → Image quality: assesses quality
- Case runner → Calibrated grading: predicts grade
- Calibrated grading → Probability calibration: calibrates probabilities
- Case runner → Grad-CAM evidence: generates explanation
- Case runner → Learned lesion channel: segments lesions
- Case runner → Classical lesion channel: detects candidates
- Case runner → Lesion evidence: builds evidence
- Case runner → Spatial agreement: checks agreement
- Case runner → Screening policy: selects disposition
- Calibrated grading → Screening policy: supplies grade
- Image quality → Screening policy: supplies quality
- Lesion evidence → Screening policy: supplies evidence
- Grad-CAM evidence → Spatial agreement: supplies heatmap
- Lesion evidence → Spatial agreement: supplies locations
- Simulation runner → District capacity model: runs simulation
- District configuration → Simulation runner: configures run