Journal of Broadcast Engineering (방송공학회논문지)

The Korean Institute of Broadcast and Media Engineers (한국방송∙미디어공학회)
권호 표지
DOI QR코드

DOI QR Code

Comparison of Performance of Medical Image Semantic Segmentation Model in ATLASV2.0 Data

ATLAS V2.0 데이터에서 의료영상 분할 모델 성능 비교

  • So Yeon Woo (Sejong University Departments of Artificial Intelligence) ;
  • Yeong Hyeon Gu (Sejong University Departments of Artificial Intelligence) ;
  • Seong Joon Yoo (Sejong University Departments of Computer Engineering)
  • Received : 2023.03.15
  • Accepted : 2023.04.05
  • Published : 2023.05.30
Download PDF
⟨ Previous Next ⟩

Abstract

There is a problem that the size of the dataset is insufficient due to the limitation of the collection of the medical image public data, so there is a possibility that the existing studies are overfitted to the public dataset. In this paper, we compare the performance of eight (Unet, X-Net, HarDNet, SegNet, PSPNet, SwinUnet, 3D-ResU-Net, UNETR) medical image semantic segmentation models to revalidate the superiority of existing models. Anatomical Tracings of Lesions After Stroke (ATLAS) V1.2, a public dataset for stroke diagnosis, is used to compare the performance of the models and the performance of the models in ATLAS V2.0. Experimental results show that most models have similar performance in V1.2 and V2.0, but X-net and 3D-ResU-Net have higher performance in V1.2 datasets. These results can be interpreted that the models may be overfitted to V1.2.

의료영상 공개 데이터는 수집에 한계가 있어 데이터셋의 양이 부족하다는 문제점이 있다. 때문에 기존 연구들은 공개 데이터셋에 과적합 되었을 우려가 있다. 본 논문은 실험을 통해 8개의 (Unet, X-Net, HarDNet, SegNet, PSPNet, SwinUnet, 3D-ResU-Net, UNETR) 의료영상 분할 모델의 성능을 비교함으로써 기존 모델의 성능을 재검증하고자 한다. 뇌졸중 진단 공개 데이터 셋인 Anatomical Tracings of Lesions After Stroke(ATLAS) V1.2과 ATLAS V2.0에서 모델들의 성능 비교 실험을 진행한다. 실험결과 대부분 모델은 V1.2과 V2.0에서 성능이 비슷한 결과를 보였다. 하지만 X-net과 3D-ResU-Net는 V1.2 데이터셋에서 더 높은 성능을 기록했다. 이러한 결과는 해당 모델들이 V1.2에 과적합 되었을 것으로 해석할 수 있다.

Keywords

Acknowledgement

본 연구는 정부(과학기술정보통신부)의 재원으로 정보통신기획평가원의 지원을 받아 수행된 연구임(No. 1711160571, 머신러닝 개발 전주기를 연결하고 쉽게 사용할 수 있는 자동화 MLOps 플랫폼 기술 개발).

References

  1. S. Zhang, S. Xu, L. Tan, H. Wang, and J. Meng, "Stroke Lesion Detection and Analysis in MRI Images Based on Deep Learning", Journal of Healthcare Engineering, vol. 2021, Apr. 2021. doi: https://doi.org/10.1155/2021/5524769
  2. S.-L. Liew et al., "A large, curated, open-source stroke neuroimaging dataset to improve lesion segmentation algorithms," Scientific Data, vol. 9, no. 1. Springer Science and Business Media LLC, Jun. 2022. doi: https://doi.org/10.1038/s41597-022-01401-7
  3. O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation," Lecture Notes in Computer Science. Springer International Publishing, pp. 234-241, October. 2015. doi : https://doi.org/10.1007/978-3-319-24574-4_28
  4. Y. Zhou, W. Huang, P. Dong, Y. Xia, and S. Wang, "D-UNet: A Dimension-Fusion U Shape Network for Chronic Stroke Lesion Segmentation," IEEE/ACM Transactions on Computational Biology and Bioinformatics, vol. 18, no. 3. Institute of Electrical and Electronics Engineers (IEEE), pp. 940-950, May. 2021. doi: https://doi.org/10.1109/tcbb.2019.2939522
  5. K. Qi et al., "X-Net: Brain Stroke Lesion Segmentation Based on Depthwise Separable Convolution and Long-Range Dependencies," Lecture Notes in Computer Science. Springer International Publishing, pp. 247-255, October. 2019. doi: https://doi.org/10.1007/978-3-030-32248-9_28
  6. V. Badrinarayanan, A. Kendall, and R. Cipolla, "SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 12. Institute of Electrical and Electronics Engineers (IEEE), pp. 2481-2495, Dec. 2017. doi: https://doi.org/10.1109/tpami.2016.2644615
  7. H. Zhao, J. Shi, X. Qi, X. Wang, and J. Jia, "Pyramid Scene Parsing Network," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Jul. 2017. doi: https://doi.org/10.1109/cvpr.2017.660
  8. P. Chao, C.-Y. Kao, Y. Ruan, C.-H. Huang, and Y.-L. Lin, "HarDNet: A Low Memory Traffic Network," 2019 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, Oct. 2019. doi: https://doi.org/10.1109/iccv.2019.00365
  9. A. Dosovitskiy et al., "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale." arXiv, 2020. doi: https://doi.org/10.48550/arXiv.2010.11929
  10. Z. Liu et al., "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows.", Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 10012-10022, 2021. doi: https://doi.org/10.48550/arXiv.2103.14030
  11. H. Cao et al., "Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation." arXiv, 2021. doi: https://doi.org/10.48550/arXiv.2105.05537
  12. N. Tomita, S. Jiang, M. E. Maeder, and S. Hassanpour, "Automatic post-stroke lesion segmentation on MR images using 3D residual convolutional neural network," NeuroImage: Clinical, vol. 27. Elsevier BV, p. 102276, 2020. doi: https://doi.org/10.1016/j.nicl.2020.102276
  13. Hatamizadeh, Ali, et al. "Unetr: Transformers for 3d medical image segmentation." Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2022. doi: https://doi.org/10.48550/arXiv.2103.10504
  14. Y. Gu, Z. Piao, and S. J. Yoo, "STHarDNet: Swin Transformer with HarDNet for MRI Segmentation," Applied Sciences, vol. 12, no. 1. MDPI AG, p. 468, Jan. 2022. doi: https://doi.org/10.3390/app12010468
  15. Anatomical Tracings of Lesions After Stroke (ATLAS) R1.1. Available online: http://fcon_1000.projects.nitrc.org/indi/retro/atlas.html (accessed on 15 Feb 2021)