Enhancing brain boundary segmentation with cm-bet: A cascaded model for high precision in brain extraction

Öz

Brain extraction is a pivotal step in neuroimaging studies, separating the brain from surrounding tissues. Accurate extraction significantly impacts subsequent analyses such as volumetric analysis, image registration, and brain parts segmentation. As in many medical fields, DL-based approaches show promising results in brain extraction studies. The success of a brain extraction model primarily depends on its effectiveness in segmenting brain boundary regions. This study proposes a 3D model with a cascade structure, CM-BET, which includes MRS-UNet as the first stage and MR-UNet as the second. The MRS-UNet generates a gray output that enhances high-frequency edge features by incorporating sharpening layers before each convolutional layer on the encoder side of the UNet. Consequently, MR-UNet, which uses the sharpened image from the first stage as input, can better focus on the brain region and produce a binary brain mask. Residual connections from ResNet50 are employed in the latent layer to preserve critical low-gradient values. To address the class imbalance across MRI slices, a hybrid loss function including focal loss is used. Evaluations of seven public datasets demonstrate that CM-BET outperforms DL-based methods in literature and widely used segmentation models such as UNet, UNet++, and DeepLabv3+. The proposed model achieved minimum-maximum dice coefficient values of 0.9809-0.9984, sensitivity values of 0.9831-0.9991, and specificity values of 0.9958-0.9992 across all datasets, outperforms DL-based studies and UNet variants.

Anahtar Kelimeler

• Brain Extraction
• Deep Learning
• Focal Loss
• Sharpening Layer
• UNet.

Proje Numarası

TEKNO-2023-038

CM-BET ile beyin sınır segmentasyonunun geliştirilmesi: Beyin çıkarımında yüksek doğruluk için kaskat bir model

Öz

Beyin çıkarımı, nörogörüntüleme çalışmalarında, beynin çevresindeki dokulardan ayrıştırılmasını sağlayan kritik bir işlemdir. Bu adımın doğruluğu; hacimsel analiz, görüntü çakıştırma, beyin bölgelerinin segmentasyonu gibi sonraki işlemlerin doğruluğunu doğrudan etkilemektedir. Tıbbın pek çok alanında olduğu gibi, derin öğrenme (DL: Deep Learning) tabanlı yaklaşımlar beyin çıkarımı görevinde de umut vadeden sonuçlar sunmaktadır. Bu süreçte, özellikle beyin sınırlarının hassas bir şekilde segmentasyonu, modellerin başarısını belirleyen en önemli faktörlerden biridir. Bu çalışma, CM-BET adını taşıyan ve kademeli yapıda tasarlanmış 3B bir DL modeli önermektedir. CM-BET, iki aşamalı bir mimariye sahiptir: İlk aşamada yer alan MRS-UNet, UNet’in kodlayıcı bloklarına entegre edilen keskinleştirme katmanları sayesinde yüksek frekanslı kenar bilgilerini öne çıkararak gri tonlamalı bir çıktı üretir. İkinci aşamadaki MR-UNet, bu keskinleştirilmiş görüntüyü girdi olarak alarak beyin bölgesini daha hassas biçimde segmentler ve ikili (binary) bir beyin maskesi üretir. Modelin latent katmanında, düşük gradyanlı beyin sınırı bilgilerini koruyabilmek amacıyla ResNet50’den alınan artık bağlantılar (residual connections) kullanılmıştır. Ayrıca, MRI dilimlerinde sıkça karşılaşılan sınıf dengesizliği sorununu ele alabilmek adına, fokal kayıp (focal loss) içeren hibrit bir kayıp fonksiyonu geliştirilmiştir. Modelin başarımı, yedi farklı kamuya açık veriseti üzerinde kapsamlı biçimde değerlendirilmiştir. Elde edilen sonuçlar, CM-BET’in literatürdeki DL tabanlı yöntemlere ve yaygın segmentasyon mimarileri olan UNet, UNet++ ve DeepLabv3+ modellerine kıyasla üstün performans sergilediğini ortaya koymaktadır. Önerilen model, tüm verisetlerinde Dice katsayısı açısından 0.9809–0.9984, duyarlılık (sensitivity) açısından 0.9831–0.9991 ve özgüllük (specificity) açısından 0.9958–0.9992 aralığında sonuçlar elde ederek hem geleneksel hem de gelişmiş UNet türevlerini geride bırakmıştır.

Anahtar Kelimeler

• Beyin çıkarımı
• derin öğrenme
• fokal kayıp
• keskinleştirme katmanı
• UNet.

Destekleyen Kurum

Sivas Cumhuriyet Üniversitesi Bilimsel Araştırma Projeleri Fonu (CUBAP)

Proje Numarası

TEKNO-2023-038

Kaynakça

1. 1. Kim, B., Kim, H., Kim, S., & Hwang, Y.R. A brief review of non-invasive brain imaging technologies and the near-infrared optical bioimaging, Applied Microscopy, 51 (1), 9, 2021.
2. 2. Akan, T., Akan, S., Alp, S., Ledbetter, C. R., Tafti, A.P., Arevalo, O., & Bhuiyan, M.A.N. Deep Learning in neuroimaging for neurodegenerative diseases: State-of-the art, Challenges, and Opportunities. Journal of the Neurological Sciences, 123735, 2025.
3. 3. Kalavathi, P., Prasath, V.S., Methods on skull stripping of MRI head scan images – a review, Journal of Digital Imaging, 29, 365–379, 2016.
4. 4. Karakis, R., Gurkahraman, K., Mitsis, G.D., Boudrias, M.H., Deep learning prediction of motor performance in stroke individuals using neuroimaging data, Journal of Biomedical Informatics, 141, 104357, 2023.
5. 5. Isensee, F., Schell, M., Pflueger, I., Brugnara, G., Bonekamp, D., Neuberger, U., Kickingereder, P., Automated brain extraction of multisequence MRI using artificial neural networks, Human Brain Mapping, 40 (17), 4952–4964, 2019.
6. 6. Shattuck, D.W., Sandor-Leahy, S.R., Schaper, K.A., Rottenberg, D.A., Leahy, R.M., Magnetic resonance image tissue classification using a partial volume model, NeuroImage, 13 (5), 856–876, 2001.
7. 7. Shattuck, D.W., Leahy, R.M., Brainsuite: an automated cortical surface identification tool, Medical Image Analysis, 6 (2), 129–142, 2002.
8. 8. Smith, S.M., Fast robust automated brain extraction. Human Brain Mapping, 17 (3), 143–155, 2002.

9. 9. Jenkinson, M., BET2: MR-based estimation of brain, skull and scalp surfaces, In: Eleventh Annu al Meeting of the Organization for Human Brain Mapping, 2005.
10. 10. Jenkinson, M., Beckmann, C.F., Behrens, T.E., Woolrich, M.W., Smith, S.M., FSL, NeuroImage, 62 (2), 782–790, 2012.
11. 11. Fischl, B., Freesurfer, NeuroImage, 62 (2), 774–781, 2012.
12. 12. Eskildsen, S.F., Coupé, P., Fonov, V., Manjón, J.V., Leung, K.K., Guizard, N., Initiative, A.D.N., BEaST: Brain extraction based on nonlocal segmentation technique, NeuroImage, 59 (3), 2362–2373, 2012.
13. 13. Souza, R., Lucena, O., Garrafa, J., Gobbi, D., Saluzzi, M., Appenzeller, S., Lotufo, R., An open, multi-vendor, multi-field-strength brain MR dataset and analysis of publicly available skull stripping methods agreement, NeuroImage, 170, 482–494, 2018.
14. 14. Ségonne, F., Dale, A.M., Busa, E., Glessner, M., Salat, D., Hahn, H.K., Fischl, B., A hybrid approach to the skull stripping problem in MRI, NeuroImage, 22 (3), 1060–1075, 2004.
15. 15. Yapici, M., Karakis, R., Gurkahraman, K., Improving brain tumor classification with deep learning using synthetic data, Computers, Materials and Continua, 74 (3), 5049–5067, 2023.
16. 16. Gurkahraman K., Karakis R., Brain tumors classification with deep learning using data augmentation, Journal of the Faculty of Engineering and Architecture of Gazi University, 36 (2), 997–1011, 2021.
17. 17. Peng, J., Wang, Y., Medical image segmentation with limited supervision: a review of deep network models, IEEE Access, 9, 36827–36851, 2021.
18. 18. Jabbar, M.A., Raoof, S.S., An insight into applications of deep learning in neuroimaging, In: Artificial Intelligence for Neurological Disorders, 323–340, Academic Press, 2023.
19. 19. Kleesiek, J., Urban, G., Hubert, A., Schwarz, D., Maier-Hein, K., Bendszus, M., Biller, A., Deep MRI brain extraction: A 3D convolutional neural network for skull stripping, NeuroImage, 129, 460–469, 2016.
20. 20. Liu, L., Cheng, J., Quan, Q., Wu, F.X., Wang, Y.P., Wang, J., A survey on U-shaped networks in medical image segmentations, Neurocomputing, 409, 244–258, 2020.
21. 21. Azad, R., Aghdam, E.K., Rauland, A., Jia, Y., Avval, A.H., Bozorgpour, A., Merhof, D., Medical image segmentation review: The success of U-Net. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46 (12), 10076–10095, 2024.
22. 22. Punn, N.S., Agarwal, S., Modality specific U-Net variants for biomedical image segmentation: A survey, Artificial Intelligence Review, 55 (7), 5845–5889, 2022.
23. 23. Hwang, H., Rehman, H.Z.U., Lee, S., 3D U-Net for skull stripping in brain MRI, Applied Sciences, 9 (3), 569, 2019. 24. Zhang, Q., Wangi, L., Zong, X., Lin, W., Li, G., Shen, D., FRNet: Flattened residual network for infant MRI skull stripping, arXiv preprint arXiv:1904.05578, 2019.
24. 25. Gurkahraman, K., Daşgın, Brain extraction from magnetic resonance images using UNet modified with residual and dense layers, Türk Doğa ve Fen Dergisi, 12 (3), 144–151, 2023.
25. 26. Hoopes, A., Mora, J.S., Dalca, A.V., Fischl, B., Hoffmann, M., SynthStrip: Skull-stripping for any brain image, NeuroImage, 260, 119474, 2022.
26. 27. Arpacı S.A., Varlı S., Improvement proposals for lesion segmentation on dermoscopic images, Journal of the Faculty of Engineering and Architecture of Gazi University, 40 (1), 251–264, 2024.
27. 28. Şahin M., Şahin E., Özdemir E., Talu F., Öztürk S., Deep learning-based automatic planning with risk minimization for brain tumor biopsy, Journal of the Faculty of Engineering and Architecture of Gazi University, 40 (1), 487–500, 2024.
28. 29. Feng, X., Wang, C., Cheng, S., Guo, L., Automatic liver and tumor segmentation of CT based on cascaded U-Net, In: Proc. Chin. Intell. Syst. Conf., 155–164, 2019.
29. 30. Chinnam, S.K.R., Sistla, V., Kolli, V.K.K., Multimodal attention-gated cascaded U-Net model for automatic brain tumor detection and segmentation, Biomedical Signal Processing and Control, 78, 103907, 2022.
30. 31. Ding, Y., Chen, F., Zhao, Y., Wu, Z., Zhang, C., Wu, D., A stacked multi-connection simple reducing net for brain tumor segmentation, IEEE Access, 7, 104011–104024, 2019.
31. 32. Gonzales, R., Woods, R.E., Digital Image Processing. Pearson Prentice Hall, USA, 2008.
32. 33. He, K., Zhang, X., Ren, S., Sun, J., Deep residual learning for image recognition, In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 770–778, 2016.
33. 34. Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollár, P., Focal loss for dense object detection, In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2980–2988, 2017.
34. 35. Huang, H., Lin, L., Tong, R., Hu, H., Zhang, Q., Iwamoto, Y., Wu, J., UNet 3+: A full-scale connected UNet for medical image segmentation, In: ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 1055–1059, 2020.
35. 36. Taha, A.A., Hanbury, A., Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool, BMC Medical Imaging, 15, 1–28, 2015.
36. 37. Puccio, B., Pooley, J.P., Pellman, J.S., Taverna, E.C., Craddock, R.C., The preprocessed connectomes project repository of manually corrected skull-stripped T1-weighted anatomical MRI data, Gigascience, 5 (1), 13742–016, 2016.
37. 38. Shattuck, D.W., Mirza, M., Adisetiyo, V., Hojatkashani, C., Salamon, G., Narr, K.L., Toga, A.W., Construction of a 3D probabilistic atlas of human cortical structures, NeuroImage, 39 (3), 1064–1080, 2008.
38. 39. Marcus, D.S., Wang, T.H., Parker, J., Csernansky, J.G., Morris, J.C., Buckner, R.L., Open Access Series of Imaging Studies (OASIS): Cross-sectional MRI data in young, middle-aged, nondemented, and demented older adults, Journal of Cognitive Neuroscience, 19 (9), 1498–1507, 2007.
39. 40. Greve, D.N., Billot, B., Cordero, D., Hoopes, A., Hoffmann, M., Dalca, A.V., Augustinack, J.C., A deep learning toolbox for automatic segmentation of subcortical limbic structures from MRI images, NeuroImage, 244, 118610, 2021.
40. 41. Dai, W., Garcia, D., De Bazelaire, C., Alsop, D.C., Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields, Magnetic Resonance in Medicine, 60 (6), 1488–1497, 2008.
41. 42. Chen, L.C., Zhu, Y., Papandreou, G., Schroff, F., Adam, H., Encoder-decoder with atrous separable convolution for semantic image segmentation, In: Proceedings of the European Conference on Computer Vision (ECCV), 801–818, 2018.
42. 43. Zhou, Z., Rahman Siddiquee, M.M., Tajbakhsh, N., Liang, J., UNet++: A nested U-Net architecture for medical image segmentation, In: Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th International Workshop, DLMIA 2018, and 8th International Workshop, ML-CDS 2018, Held in Conjunction with MICCAI 2018, 3–11, 2018.