EdgeCVS

Models from EdgeCVS: Democratization of Surgical AI with a Distilled Edge-Deployable Critical View of Safety (CVS) Model (MICCAI 2026).

EdgeCVS assesses the three Critical View of Safety criteria in laparoscopic cholecystectomy from single frames, with a small model distilled from large multi-task teachers that runs in real time on a CPU.

Code: github.com/IMSY-DKFZ/edgecvs

Models

Each model lives in its own folder, named like the code's exp option:

Model (exp)	Backbone	Params	Weights	SAGES-CVS 2024 test mAP
edgecvs-5m	EdgeNeXt-Small	5.6M	fp16, 11 MB	66.4

edgecvs-5m/
├── config.yaml                     model configuration used to build the network
└── edgecvs-5m-fold0.safetensors    weights (fold 0, fp16; loaded into an fp32 model)

EdgeCVS-5M

• Input: RGB frame, resized to 448×448, ImageNet normalization
• Output: probabilities for C1 (two structures), C2 (hepatocystic triangle) and C3 (cystic plate); optional 7-class anatomy segmentation (background, cystic plate, hepatocystic triangle, cystic artery, cystic duct, gallbladder, tool)
• Deployment: the segmentation head is optional; the classifier alone is used for the CSV predictions and the speed figures below

License

CC BY-NC 4.0: the models are trained on data released for non-commercial use.

Citation

@InProceedings{YamAmi_EdgeCVS_MICCAI2026,
        author = { Yamlahi, Amine AND Hennighausen, Jakob AND Hansen, Pascal AND Leeb, David AND Maier-Hein, Lena},
        title = { { EdgeCVS: Democratization of surgical AI with a Distilled Edge-Deployable Critical View of Safety (CVS) model } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 16892},
        month = {September},
        page = {pending}
}