README.md

metadata

license: apache-2.0
library_name: pytorch
pipeline_tag: image-feature-extraction
tags:
  - medical-imaging
  - ct
  - abdominal-ct
  - radiology
  - vision-transformer
  - dinov2
  - pytorch
  - feature-extraction

AbdomenNet

Public release assets for:

Acute abdomen on non-contrast CT: a foundation model for diagnosis, risk stratification and emergency triage

Files

• teacher_checkpoint.pth
  • released pretraining teacher checkpoint for AbdomenNet / AbCT

Checkpoint format

teacher_checkpoint.pth is a backbone-style release checkpoint.

• top-level key: teacher
• no optimizer state
• no scheduler state
• no epoch / scaler / results payload

Minimal loading example

import torch

ckpt = torch.load("teacher_checkpoint.pth", map_location="cpu")
teacher_state = ckpt["teacher"]
backbone_state = {
    k.replace("backbone.", "", 1): v
    for k, v in teacher_state.items()
    if k.startswith("backbone.")
}
print(len(backbone_state))

Code

Source code and training / fine-tuning details:

• https://anonymous.4open.science/r/Acute_abdomen/readme.md

Notes

• This repository only hosts the released checkpoint.
• Large qualitative test-case bundles are not included in the default public release.