Cross-Modality Domain Adaptation for Medical Image Segmentation

Domain Adaptation (DA) has recently raised strong interest in the medical imaging community. By encouraging algorithms to be robust to unseen situations or different input data domains, Domain Adaptation improves the applicability of machine learning approaches to various clinical settings. While a large variety of DA techniques has been proposed, most of these techniques have been validated either on private datasets [4,5] or on small publicly available datasets [6,7,8,9]. Moreover, these datasets mostly address single-class problems. To tackle these limitations, the crossMoDA challenge introduced the first large and multi-class dataset for unsupervised cross-modality Domain Adaptation. From an application perspective, crossMoDA focuses on MRI segmentation for Vestibular Schwannoma. Compared to the previous crossMoDA instance, which made use of multiinstitutional data acquired in controlled conditions for radiosurgery planning and focused on a 2 class segmentation task (tumour and cochlea), the 2023 edition extends the segmentation task by including multi-institutional, heterogenous data acquired for routine surveillance purposes and introduces a sub-segmentation for the tumour (intra- and extrameatal components) thereby leading to a 3 class problem. The goal of the segmentation task is to segment two key brain structures (tumour and cochlea) involved in the follow-up and treatment planning of vestibular schwannoma (VS). The segmentation of these two structures is required for radiosurgery, a common VS treatment, and is also key to improve routine surveillance. Anatomically, the tumour region can be divided into two sub regions, called intra- and extra-meatal regions, corresponding to being inside or outside the inner ear canal respectively. The guidelines for reporting results in vestibular schwannoma mention that the intra-meatal and extra-meatal portions of the tumour must be distinguished, and the largest extra-meatal diameter should be used to report the size of the tumour [18]. Moreover, the size and volume features extracted from the extra-meatal region are considered as the most sensitive radiomic features for the evaluation of VS growth [19]. For crossMoDA 2023, we introduce for the first time the segmentation of the tumour split into the intra- and extra-meatal regions. Currently, the diagnosis and surveillance of patients with VS are commonly performed using contrast-enhanced T1 (ceT1) MR imaging. However, there is growing interest in using noncontrast imaging sequences such as high-resolution T2 imaging, as it mitigates the risks associated with gadolinium-containing contrast agents [1,2]. In addition to improving patient safety, T2 imaging is 10 times more cost-efficient than ceT1 imaging [16]. For this reason, we initially proposed and continue to focus on an unsupervised cross-modality segmentation benchmark (from ceT1 to T2) that aims to automatically perform VS (including intra-/extrameatal sub-segmentation) and cochlea segmentation on T2 scans. The training source and target sets are respectively unpaired annotated ceT1 and non-annotated T2 scans. To validate the robustness of the proposed approaches on different T2 settings, multiinstitutional, heterogeneous scans from centres in the UK and Tilburg, NL are used in this task. [1] Shapey, J., et al: An artificial intelligence framework for automatic segmentation and volumetry of Vestibular Schwannomas from contrast-enhanced t1-weighted and high-resolution t2-weighted MRI. In: Journal of Neurosurgery JNS. (2019) [2] Wang, G., et al: Automatic segmentation of Vestibular Schwannoma from T2-weighted MRI by deep spatial attention with hardness-weighted loss. In: MICCAI 2019. (2019) [3] Van de Langenberg, R., et al: Follow-up assessment of vestibular schwannomas: volume quantification versus two-dimensional measurements. In: Neuroradiology 51, 517 (2009). [4] Kamnitsas, K., et al: Unsupervised domain adaptation in brain lesion segmentation with adversarial networks. In: IPMI (2017) [5] Yang, J., et al: Unsupervised Domain Adaptation via Disentangled Representations: Application to Cross- Modality Liver Segmentation. In: MICCAI 2019. (2019) [6] Chen, C., el al: Synergistic Image and Feature Adaptation: Towards Cross-Modality Domain Adaptation for Medical Image Segmentation. In: AAAI-19. (2019) [7] Dou, Q., et al: Pnp-adanet: Plug-and-play adversarial domain adaptation network with a benchmark at cross- modality cardiac segmentation. ArXiv. (2018) [8] Orbes-Arteaga, et al: Multi-domain adaptation in brain MRI through paired consistency and adversarial learning. In: DART - Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data. (2019) [9] Xue, Y., el al: Dual-task Self-supervision for Cross-Modality Domain Adaptation. In: MICCAI 2020. (2020) [10] Maier-Hein, L., Eisenmann, M., et al: “Is the winner really the best? A critical analysis of common research practice in biomedical image analysis competitions”. ArXiv:1806.02051, (2018) [11] Bakas, S., et al: "Identifying the Best Machine Learning Algorithms for Brain Tumor Segmentation, Progression Assessment, and Overall Survival Prediction in the BRATS Challenge". ArXiv:1811.02629 (2018). [12] Maier, O., et al: "ISLES 2015 - A public evaluation benchmark for ischemic stroke lesion segmentation from multispectral MRI". Medical Image Analysis (2017). [13] Koos WT, et al: "Neurotopographic considerations in the microsurgical treatment of small acoustic neurinomas". Journal of Neurosurgery (1998) [14] Dale, A.M., et al: "Cortical surface-based analysis. I. Segmentation and surface reconstruction". Neuroimage (1999). [15] Cardoso, J., et al: "Geodesic Information Flows: Spatially-Variant Graphs and Their Application to Segmentation and Fusion". Transactions on Medical Imaging (2015) [16] Coelho, D, et al: "MRI surveillance of vestibular schwannomas without contrast enhancement: Clinical and economic evaluation". Laryngoscope (2018) [17] Baccianella, S, et al: "Evaluation Measures for Ordinal Regression". International Conference on Intelligent Systems Design and Applications (ISDA) (2009) [18] Kanzaki, Jin, et al. "New and modified reporting systems from the consensus meeting on systems for reporting results in vestibular schwannoma." Otology & neurotology 24.4 (2003): 642-649. [19] Connor, Steve EJ. "Imaging of the vestibular schwannoma: diagnosis, monitoring, and treatment planning." Neuroimaging Clinics 31.4 (2021): 451-471. [20] Wijethilake, Navodini, et al. "Boundary Distance Loss for Intra-/Extra-meatal Segmentation of Vestibular Schwannoma." International Workshop on Machine Learning in Clinical Neuroimaging. Springer, Cham, 2022. [21] Kujawa, Aaron, et al. "Deep Learning for Automatic Segmentation of Vestibular Schwannoma: A Retrospective Study from Multi-Centre Routine MRI." medRxiv (2022). [22] Dorent, Reuben, et al. "CrossMoDA 2021 challenge: Benchmark of cross-modality domain adaptation techniques for vestibular schwannoma and cochlea segmentation." Medical Image Analysis 83 (2023): 102628