연세대학교 DELAB Vision 팀에서 진행한 발표 자료입니다.

Medical Segmentation Dataset인 BRATS, BTCV와
Segmentation의 평가지표인 Dice Score, Hausdorff Distance에 대한 설명도 추가하였습니다.

혹시 피드백이나 문제점이 있으면 알려주시면 감사하겠습니다!

– 논문: https://arxiv.org/pdf/2201.01266.pdf
– 코드(BraTS): https://github.com/Project-MONAI/research-contributions/tree/main/SwinUNETR/BRATS21

– 논문: https://openaccess.thecvf.com/content/CVPR2022/papers/Tang_Self-Supervised_Pre-Training_of_Swin_Transformers_for_3D_Medical_Image_Analysis_CVPR_2022_paper.pdf
– 코드(BTCV): https://github.com/Project-MONAI/research-contributions/tree/main/SwinUNETR/BTCV
– 코드(pretrain): https://github.com/Project-MONAI/research-contributions/tree/main/SwinUNETR/Pretrain

Swin UNETR

<PT 자료>

<정리(발표) 자료>

Self Supervised Pre-training of Swin Transformer

<참고 자료>

1. BraTS 2021
http://braintumorsegmentation.org/

2. Dice Score
http://machinelearningkorea.com/2019/07/13/%ED%8F%89%EA%B0%80%EC%A7%80%ED%91%9C-dice/

3. Hausdorff Distance
3-1. https://structseg2019.grand-challenge.org/Evaluation/
3-2. https://npclinic3.tistory.com/7
3-3. https://dhpark1212.tistory.com/entry/Hausdorff-Distance

4. Normalization
[AI] 그림으로 보는 normalization 기법 (tistory.com)