nnmamba: 3d biomedical image segmentation, classification and landmark detection with state space model

Haifan Gong, Luoyao Kang, Yitao Wang, Yihan WANG, Xiang Wan, Xusheng Wu, Haofeng Li

Abstract:

n biomedical image analysis, developing architectures thateffectively capture long-range dependencies is crucial. Tra-ditional Convolutional Neural Networks (CNNs) are con-strained by their local receptive fields, while Transformers,though proficient in global context integration, are compu-tationally demanding for high-dimensional medical images.Here, we present nnMamba, a novel architecture that com-bines the strengths of CNNs with the long-range modelingcapabilities of State Space Models (SSMs). We introduce theMamba-In-Convolution with Channel-Spatial Siamese learn-ing (MICCSS) block to model long-range voxel relationships.Additionally, we implement channel scaling and channel-sequential learning methods to enhance performance in denseprediction and classification tasks. Extensive experimentson seven datasets demonstrate that nnMamba outperformscurrent state-of-the-art methods in 3D image segmentation,classification, and landmark detection. nnMamba effectivelyintegrates CNNs’ local representation with SSMs’ global con-text processing, establishing a new benchmark for long-rangedependency modeling in medical image analysis. Code isavailable at https://github.com/lhaof/nnMamba

(PDF) Nnmamba: 3D Biomedical Image Segmentation, Classification and Landmark Detection with State Space Model. Available from: https://www.researchgate.net/publication/391694473_Nnmamba_3D_Biomedical_Image_Segmentation_Classification_and_Landmark_Detection_with_State_Space_Model#fullTextFileContent [accessed Nov 12 2025].