# AOCR2024 AI Challenge — Data

来源：https://www.kaggle.com/competitions/aocr2024/data

采集时间：2026-10-18T13:43:06Z

Far Eastern Memorial Hospital in New Taipei City, Taiwan has provided the image data and annotations. From this collection, we've chosen contrast-enhanced abdominal and pelvic CT scans from adults over 18 years of age. These scans, taken in the axial section, have a slice thickness of 5mm and cover a span of the last ten years.

- The training dataset comprises 500 CT scans from patients confirmed with acute appendicitis and an additional 500 scans from individuals who exhibited acute abdominal pain but were diagnosed without appendicitis. We've pre-divided the data into Train and Validation sets; adhering to this suggested split is advised.
- Test data: 200 CT scans from acute abdomen condition, but the composition ratio is unknown.
- We've created mask files for the CT scans showing acute appendicitis by delineating the affected appendix, as illustrated below in NIfTI format. Annotations were focused on the abnormal appendices, and inflammation in adjacent areas was not marked. We have taken extensive precautions to ensure the accuracy of both the images and annotations. However, in the unlikely event of a minor annotation error, we believe it will not compromise the fairness of the competition.

![](assets/693c6b0337cd78d1280eb17c.png)

![](assets/c7b0446d6bfd6c159aba164c.jpg)


**Evaluation**

- Each CT slice is evaluated to identify the presence of an abnormal appendix indicative of appendicitis. Furthermore, the entire CT scan is assessed to determine a diagnosis of acute appendicitis for that particular scan session. The final score combines evaluations from both slice-level and scan-level assessments.

![](assets/b960196a3723988bb985bd8a.jpg)