L3 CT Body Composition & Sarcopenia (ONNX)
2D single-slice L3 axial CT segmentation models exported from nnU-Net v2 to ONNX (opset 17, dynamic INT8 quantized) for client-side browser and edge inference.
Models
| File | Cohort / Source | Classes | Patch Size | Target Spacing | Size |
|---|---|---|---|---|---|
pgimer_l3_int8.onnx |
PGIMER Indian Cohort (Gupta et al., Indian J Gastroenterol 2026, PMID: 41553453) | 2 (0: background, 1: skeletal muscle) |
512 x 512 |
0.78125 x 0.78125 mm |
55.4 MB |
mazurowski_l3_int8.onnx |
Mazurowski Lab (CT-Muscle-and-Fat-Segmentation, CC BY-NC 4.0) |
5 (0: bg, 1: muscle, 2: SAT, 3: VAT, 4: IMAT) |
256 x 256 |
1.0 x 1.0 mm |
109.5 MB |
See models.json for exact CTNormalization parameters (clipMin, clipMax, mean, std) and SHA-256 checksums.
Live Browser App
Available at https://medstats.github.io/sarcopenia/
Created by Anupam Kumar Singh, Professor of Medicine, Santosh Medical College.
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support