Learning Generalizable 3D Medical Image Representations from Mask-Guided Self-Supervision

• URL: http://arxiv.org/abs/2603.13660v1
• Date: Sat, 14 Mar 2026 00:06:18 GMT
• Title: Learning Generalizable 3D Medical Image Representations from Mask-Guided Self-Supervision
• Abstract summary: MASS (MAsk-guided Self-Supervised learning) treats in-context segmentation as the pretext task for learning general-purpose medical imaging representations.<n>By training on thousands of diverse mask proposals spanning anatomical structures and pathological findings, MASS learns what semantically defines medical structures.
• Score: 24.431507546276638
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Foundation models have transformed vision and language by learning general-purpose representations from large-scale unlabeled data, yet 3D medical imaging lacks analogous approaches. Existing self-supervised methods rely on low-level reconstruction or contrastive objectives that fail to capture the anatomical semantics critical for medical image analysis, limiting transfer to downstream tasks. We present MASS (MAsk-guided Self-Supervised learning), which treats in-context segmentation as the pretext task for learning general-purpose medical imaging representations. MASS's key insight is that automatically generated class-agnostic masks provide sufficient structural supervision for learning semantically rich representations. By training on thousands of diverse mask proposals spanning anatomical structures and pathological findings, MASS learns what semantically defines medical structures: the holistic combination of appearance, shape, spatial context, and anatomical relationships. We demonstrate effectiveness across data regimes: from small-scale pretraining on individual datasets (20-200 scans) to large-scale multi-modal pretraining on 5K CT, MRI, and PET volumes, all without annotations. MASS demonstrates: (i) few-shot segmentation on novel structures, (ii) matching full supervision with only 20-40\% labeled data while outperforming self-supervised baselines by over 20 in Dice score in low-data regimes, and (iii) frozen-encoder classification on unseen pathologies that matches full supervised training with thousands of samples. Mask-guided self-supervised pretraining captures broadly generalizable knowledge, opening a path toward 3D medical imaging foundation models without expert annotations. Code is available: https://github.com/Stanford-AIMI/MASS.

Related papers

• EndoVLM: An Endoscopy Vision-Language Pre-training Model via Anatomy-Guided Sparsity and Progressive Alignment [15.371433625713232]
  EndoVLM is a novel vision-language FM pre-trained on over 348K endoscopic examinations.<n>An Anatomy-Guided Sparse Pooling mechanism utilizes textual descriptions as queries to drive sparse attention.<n>A Progressive Semantic-Aware Alignment strategy models clinical taxonomy (anatomy and pathological status) via structured soft targets.<n>A Semantic-Concentrated Masked Autoencoder is applied exclusively to these semantic-rich frames, integrating low-level visual precision with robust high-level semantic representation.
  arXiv  Detail & Related papers  (2026-08-05T05:56:41Z)
• M3Ret: Unleashing Zero-shot Multimodal Medical Image Retrieval via Self-Supervision [24.846428105192405]
  We train M3Ret, a unified visual encoder, without any modality-specific customization.<n>It successfully learns transferable representations using both generative (MAE) and contrastive (SimDINO) self-supervised learning (SSL) paradigms.<n>Our approach sets a new state-of-the-art in zero-shot image-to-image retrieval across all individual modalities, surpassing strong baselines such as DINOv3 and the text-supervised BMC-CLIP.
  arXiv  Detail & Related papers  (2025-09-01T10:59:39Z)
• Autoregressive Sequence Modeling for 3D Medical Image Representation [48.706230961589924]
  We introduce a pioneering method for learning 3D medical image representations through an autoregressive sequence pre-training framework.<n>Our approach various 3D medical images based on spatial, contrast, and semantic correlations, treating them as interconnected visual tokens within a token sequence.
  arXiv  Detail & Related papers  (2024-09-13T10:19:10Z)
• MiM: Mask in Mask Self-Supervised Pre-Training for 3D Medical Image Analysis [9.472502717128556]
  Masked AutoEncoder (MAE) for feature pre-training can unleash the potential of ViT on various medical vision tasks.<n>We propose a novel textitMask in Mask (MiM) pre-training framework for 3D medical images.
  arXiv  Detail & Related papers  (2024-04-24T01:14:33Z)
• MASSM: An End-to-End Deep Learning Framework for Multi-Anatomy Statistical Shape Modeling Directly From Images [1.9029890402585894]
  We introduce MASSM, a novel end-to-end deep learning framework that simultaneously localizes multiple anatomies, estimates population-level statistical representations, and delineates shape representations directly in image space. Our results show that MASSM, which delineates anatomy in image space and handles multiple anatomies through a multitask network, provides superior shape information compared to segmentation networks for medical imaging tasks.
  arXiv  Detail & Related papers  (2024-03-16T20:16:37Z)
• Disruptive Autoencoders: Leveraging Low-level features for 3D Medical Image Pre-training [51.16994853817024]
  This work focuses on designing an effective pre-training framework for 3D radiology images. We introduce Disruptive Autoencoders, a pre-training framework that attempts to reconstruct the original image from disruptions created by a combination of local masking and low-level perturbations. The proposed pre-training framework is tested across multiple downstream tasks and achieves state-of-the-art performance.
  arXiv  Detail & Related papers  (2023-07-31T17:59:42Z)
• LVM-Med: Learning Large-Scale Self-Supervised Vision Models for Medical Imaging via Second-order Graph Matching [59.01894976615714]
  We introduce LVM-Med, the first family of deep networks trained on large-scale medical datasets. We have collected approximately 1.3 million medical images from 55 publicly available datasets. LVM-Med empirically outperforms a number of state-of-the-art supervised, self-supervised, and foundation models.
  arXiv  Detail & Related papers  (2023-06-20T22:21:34Z)
• Mine yOur owN Anatomy: Revisiting Medical Image Segmentation with Extremely Limited Labels [54.58539616385138]
  We introduce a novel semi-supervised 2D medical image segmentation framework termed Mine yOur owN Anatomy (MONA) First, prior work argues that every pixel equally matters to the model training; we observe empirically that this alone is unlikely to define meaningful anatomical features. Second, we construct a set of objectives that encourage the model to be capable of decomposing medical images into a collection of anatomical features.
  arXiv  Detail & Related papers  (2022-09-27T15:50:31Z)
• CUTS: A Deep Learning and Topological Framework for Multigranular Unsupervised Medical Image Segmentation [8.307551496968156]
  We present CUTS, an unsupervised deep learning framework for medical image segmentation. For each image, it produces an embedding map via intra-image contrastive learning and local patch reconstruction. CUTS yields a series of coarse-to-fine-grained segmentations that highlight features at various granularities.
  arXiv  Detail & Related papers  (2022-09-23T01:09:06Z)
• Intelligent Masking: Deep Q-Learning for Context Encoding in Medical Image Analysis [48.02011627390706]
  We develop a novel self-supervised approach that occludes targeted regions to improve the pre-training procedure. We show that training the agent against the prediction model can significantly improve the semantic features extracted for downstream classification tasks.
  arXiv  Detail & Related papers  (2022-03-25T19:05:06Z)
• Few-shot Medical Image Segmentation using a Global Correlation Network with Discriminative Embedding [60.89561661441736]
  We propose a novel method for few-shot medical image segmentation. We construct our few-shot image segmentor using a deep convolutional network trained episodically. We enhance discriminability of deep embedding to encourage clustering of the feature domains of the same class.
  arXiv  Detail & Related papers  (2020-12-10T04:01:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.