Existing approaches to 3D medical image segmentation can be generally categorized into convolution-based or transformer-based methods. While convolutional neural networks (CNNs) demonstrate proficiency in extracting local features, they encounter challenges in capturing global representations. In contrast, the consecutive self-attention modules present in vision transformers excel at capturing long-range dependencies and achieving an expanded receptive field. In this paper, we propose a novel approach, termed SCANeXt, for 3D medical image segmentation. Our method combines the strengths of dual attention (Spatial and Channel Attention) and ConvNeXt to enhance representation learning for 3D medical images. In particular, we propose a novel self-attention mechanism crafted to encompass spatial and channel relationships throughout the entire feature dimension. To further extract multiscale features, we introduce a depth-wise convolution block inspired by ConvNeXt after the dual attention block. Extensive evaluations on three benchmark datasets, namely Synapse, BraTS, and ACDC, demonstrate the effectiveness of our proposed method in terms of accuracy. Our SCANeXt model achieves a state-of-the-art result with a Dice Similarity Score of 95.18% on the ACDC dataset, significantly outperforming current methods.

Cervical radiculopathy greatly affects patients’ quality of life with symptoms of upper limb pain or motor dysfunctions [ 1 x [1] Carette, S and Fehlings, MG. Cervical radiculopathy. N Engl J Med. 2005; 353: 392–399 https://doi.org/10.1056/NEJMcp043887 Crossref | PubMed | Scopus (185) | Google Scholar See all References ] [1] . The annual incidence was reported to be 107.3 per 10,000 for men and 63.5 per 10,000 for women [ 2 x [2] Woods, BI and Hilibrand, AS. Cervical radiculopathy: epidemiology, etiology, diagnosis, and treatment. J Spinal Disord Tech. 2015; 28: E251–E259 https://doi.org/10.1097/BSD.0000000000000284 Crossref | PubMed | Scopus (29) | Google Scholar See all References ] [2] . It has been shown that neurogenic inflammation induced by mechanical compression of cervical spinal nerve root is the major culprit contributing to radicular pain [ 1 x [1] Carette, S and Fehlings, MG. Cervical radiculopathy. N Engl J Med. 2005; 353: 392–399 https://doi.org/10.1056/NEJMcp043887 Crossref | PubMed | Scopus (185) | Google Scholar See all References ] [1] . Currently, nonsteroid anti-inflammatory drugs, opioids, and steroids are prescribed as conservative treatments [ 1 x [1] Carette, S and Fehlings, MG. Cervical radiculopathy. N Engl J Med. 2005; 353: 392–399 https://doi.org/10.1056/NEJMcp043887 Crossref | PubMed | Scopus (185) | Google Scholar See all References , 3 x [3] DeLeo, JA, Hashizume, H, Rutkowski, MD, and Weinstein, JN. Cyclooxygenase-2 inhibitor SC-236 attenuates mechanical allodynia following nerve root injury in rats. J Orthop Res. 2000; 18: 977–982 https://doi.org/10.1002/jor.1100180618 Crossref | PubMed | Scopus (17) | Google Scholar See all References , 4 x [4] Sekiguchi, M, Sekiguchi, Y, Konno, S, Kobayashi, H, Homma, Y, and Kikuchi, S. Comparison of neuropathic pain and neuronal apoptosis following nerve root or spinal nerve compression. Eur Spine J. 2009; 18: 1978–1985 https://doi.org/10.1007/s00586-009-1064-z Crossref | PubMed | Scopus (48) | Google Scholar See all References , 5 x [5] Slotkin JR, Ness JK, Snyder KM, Skiles AA, Woodard EJ, OʼShea T, et al. Sustained local release of methylprednisolone from a thiol-acrylate poly(ethylene glycol) hydrogel for treating chronic compressive radicular pain: Spine 2015:1. doi:10.1097/BRS.0000000000001309. Google Scholar See all References ]. However, pain often remains insufficiently controlled and further investigations aiming better treatment approaches is warranted [ 6 x [6] Liu, D-L, Wang, X, Chu, W-G, Lu, N, Han, W-J, Du, Y-K et al. Chronic cervical radiculopathic pain is associated with increased excitability and hyperpolarization-activated current (Ih) in large-diameter dorsal root ganglion neurons. Mol Pain. 2017; 13: 1744806917707127 https://doi.org/10.1177/1744806917707127 Crossref | Scopus (1) | Google Scholar See all References , 7 x [7] Rosenberg, GJ, Yee, AJM, and Erwin, WM. Bedside to bench and back to bedside: translational implications of targeted intervertebral disc therapeutics. J Orthop Transl. 2017; 10: 18–27 https://doi.org/10.1016/j.jot.2017.03.008 Crossref | PubMed | Scopus (0) | Google Scholar See all References ].

Data augmentation methods inspired by CutMix have demonstrated significant potential in recent semi-supervised medical image segmentation tasks. However, these approaches often apply CutMix operations in a rigid and inflexible manner, while paying insufficient attention to feature-level consistency constraints. In this paper, we propose a novel method called Mutual Mask Mix with High-Low level feature consistency (M$^3$HL) to address the aforementioned challenges, which consists of two key components: 1) M$^3$: An enhanced data augmentation operation inspired by the masking strategy from Masked Image Modeling (MIM), which advances conventional CutMix through dynamically adjustable masks to generate spatially complementary image pairs for collaborative training, thereby enabling effective information fusion between labeled and unlabeled images. 2) HL: A hierarchical consistency regularization framework that enforces high-level and low-level feature consistency between unlabeled and mixed images, enabling the model to better capture discriminative feature representations.Our method achieves state-of-the-art performance on widely adopted medical image segmentation benchmarks including the ACDC and LA datasets. Source code is available at https://github.com/PHPJava666/M3HL