Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fake It Right: Injecting Anatomical Logic into Synthetic Supervised Pre-training for Medical Segmentation

Jiaqi Tang, Mengyan Zheng, Shu Zhang, Fandong Zhang, Qingchao Chen

TL;DR

An Anatomy-Informed Synthetic Supervised Pre-training framework unifying FDSL's infinite scalability with anatomical realism is proposed unifying FDSL's infinite scalability with anatomical realism.

Abstract

Vision Transformers (ViTs) excel in 3D medical segmentation but require massive annotated datasets. While Self-Supervised Learning (SSL) mitigates this using unlabeled data, it still faces strict privacy and logistical barriers. Formula-Driven Supervised Learning (FDSL) offers a privacy-preserving alternative by pre-training on synthetic mathematical primitives. However, a critical semantic gap limits its efficacy: generic shapes lack the morphological fidelity, fixed spatial layouts, and inter-organ relationships of real anatomy, preventing models from learning essential global structural priors. To bridge this gap, we propose an Anatomy-Informed Synthetic Supervised Pre-training framework unifying FDSL's infinite scalability with anatomical realism. We replace basic primitives with a lightweight shape bank with de-identified, label-only segmentation masks from 5 subjects. Furthermore, we introduce a structure-aware sequential placement strategy to govern the patch synthesis process. Instead of random placement, we enforce physiological plausibility using spatial anchors for correct localization and a topological graph to manage inter-organ interactions (e.g., preventing impossible overlaps). Extensive experiments on BTCV and MSD datasets demonstrate that our method significantly outperforms state-of-the-art FDSL baselines and SSL methods by 1.74\% and up to 1.66\%, while exhibiting a robust scaling effect where performance improves with increased synthetic data volume. This provides a data-efficient, privacy-compliant solution for medical segmentation. The code will be made publicly available upon acceptance.

Fake It Right: Injecting Anatomical Logic into Synthetic Supervised Pre-training for Medical Segmentation

TL;DR

An Anatomy-Informed Synthetic Supervised Pre-training framework unifying FDSL's infinite scalability with anatomical realism is proposed unifying FDSL's infinite scalability with anatomical realism.

Abstract

Vision Transformers (ViTs) excel in 3D medical segmentation but require massive annotated datasets. While Self-Supervised Learning (SSL) mitigates this using unlabeled data, it still faces strict privacy and logistical barriers. Formula-Driven Supervised Learning (FDSL) offers a privacy-preserving alternative by pre-training on synthetic mathematical primitives. However, a critical semantic gap limits its efficacy: generic shapes lack the morphological fidelity, fixed spatial layouts, and inter-organ relationships of real anatomy, preventing models from learning essential global structural priors. To bridge this gap, we propose an Anatomy-Informed Synthetic Supervised Pre-training framework unifying FDSL's infinite scalability with anatomical realism. We replace basic primitives with a lightweight shape bank with de-identified, label-only segmentation masks from 5 subjects. Furthermore, we introduce a structure-aware sequential placement strategy to govern the patch synthesis process. Instead of random placement, we enforce physiological plausibility using spatial anchors for correct localization and a topological graph to manage inter-organ interactions (e.g., preventing impossible overlaps). Extensive experiments on BTCV and MSD datasets demonstrate that our method significantly outperforms state-of-the-art FDSL baselines and SSL methods by 1.74\% and up to 1.66\%, while exhibiting a robust scaling effect where performance improves with increased synthetic data volume. This provides a data-efficient, privacy-compliant solution for medical segmentation. The code will be made publicly available upon acceptance.
Paper Structure (9 sections, 3 equations, 2 figures, 3 tables)

This paper contains 9 sections, 3 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Theoretical Formulation
  4. Anatomy-Informed Shape Generation
  5. Structure-Aware Sequential Placement
  6. Experiments
  7. Experimental Setup
  8. Results and Analysis
  9. Conclusion

Figures (2)

  • Figure 1: Overall framework. By extracting morphological and spatial priors from minimal real masks, our generator iteratively synthesizes unlimited, biologically plausible data to bridge the semantic gap of FDSL for robust model pre-training and downstream fine-tuning.
  • Figure 2: Left: Qualitative comparison results. Red dashes indicate misidentified areas and blue dashes indicate more accurately identified areas. Right: Scaling effect when controlling the generated data number from 500 to 50000.