Enhancing Medical Lesion Segmentation via Several Simple Modules

Ruichen Li, Zhao YaoarXiv 2026 · arXiv

Medical image segmentation is the backbone of automated clinical diagnosis, yet it remains a "high-stakes, low-data" challenge. While deep learning has made massive strides, models often struggle with rare pathologies where data is scarce, or they suffer from "catastrophic forgetting" when asked to learn new classes. Even when they achieve high accuracy scores, the resulting masks can be morphologically nonsensical—think "hollow" tumors or fragmented predictions.

A recent paper from researchers at Hunan University, titled "Enhancing Medical Lesion Segmentation via Several Simple Modules," proposes a trio of lightweight solutions to these persistent problems. Instead of building a massive new architecture, they introduce three plug-and-play modules designed to improve topological consistency, cross-domain generalization, and semantic robustness.

TL;DR

The authors introduce three lightweight modules—a differentiable morphological loss, a 3D volumetric prompting system, and a synonym-aware segmentation head—to fix common failures in medical AI. These modules significantly reduce morphological distortions, enable faster training, and allow models to handle inconsistent clinical terminology (synonyms) without losing performance.

What Problem Does This Paper Address?

Current segmentation models face three major hurdles:

  1. Morphological Fragility: Standard loss functions (like Dice loss) treat every voxel independently. This often results in "broken" or "hollow" segmentation masks for small, complex lesions because the model doesn't "understand" that a tumor should be a continuous, solid shape.
  2. The 2D-to-3D Gap: Many efficient models process medical scans slice-by-slice (2D). This ignores the inter-slice continuity of human anatomy, leading to "false positives" where a model predicts a lesion in a slice where it shouldn't exist.
  3. Semantic Ambiguity: In clinical practice, the same pathology might be called different things (e.g., "pancreatic cancer" vs. "pancreatic malignant neoplasm"). Standard models treat these as entirely different categories, leading to "gradient scattering" and poor generalization.

Methods

The researchers propose three specific "enhancements":

1. DT-Morph Combined Loss

To fix broken shapes, the authors developed a Differentiable Distance-Transform Morphological Loss. They simulate "morphological closing" (a classic image processing technique to fill holes and bridge gaps) using continuous spatial pooling. Because this process is now differentiable, it can be baked directly into the neural network's training loop.

Fig. 1. Overview of DT-MorphLoss. The overall pipeline proceeds as follows: 1) a Sigmoid activation and Otsu binarization convert the prediction map into a binary mask; 2) the mask is fed in parallel into two branches, where the morphological-loss branch applies erosion and dilation to compute Lmorph, while the distance-transform branch extracts an edge mask to produce a distance field for Ldt.; 3) the two terms are weighted and combined with the standard Dice loss (Ldice) and cross-entropy loss (Lce) to obtain the total loss.

2. 3D Group Prompt (GroupPrompt3D)

To move beyond 2D limitations, they introduced a volumetric prompting mechanism. By using a "support volume" (a reference image with a known mask) to guide the "query volume" (the scan being analyzed), the model gains 3D spatial awareness. They use a lightweight 3D ASPP (Atrous Spatial Pyramid Pooling) that cuts 3D computation costs by over 90% compared to standard versions.

3. Many-to-One Conditional Segmentation Head (CSH)

This module addresses the "synonym" problem. It uses a hypernetwork to generate parameters on the fly. Crucially, it includes a mapping layer that uses SHA-1 hashing to route various clinical synonyms to a single "canonical" category. This prevents the model from getting confused by different names for the same thing.

Fig. 3. Overview of Many-to-one-CSH. Visual queries are encoded into multi-level fea- tures, while the text branch uses a Hash function to map various input texts into cor- responding targets within a vocabulary registry and generate specific prompts. These prompts are transformed into distinctive text embeddings ωi, which are sequentially combined with the visual features and fed into a parameter generator composed of mul- tiple specialized MLP heads. This module dynamically generates prompt-conditioned weight parameters θi that directly modulate the layers within the segment head, al- lowing the decoded query features to be processed into final masks.

Datasets

The modules were evaluated across four major CT datasets:

  • MSWAL: An abdominal tumor dataset (7 labels) used as the primary benchmark.
  • AbdomenAtlas 3.0: A large-scale atlas focusing on four specific tumor types.
  • AMOS22: A multi-organ benchmark with 16 anatomical labels.
  • FLARE22: Another multi-organ dataset with 15 anatomical labels.

Key Insights and Ideas

  • Topology Matters More Than Pixels: By forcing the model to respect "anatomical connectivity" through the DT-Morph loss, the researchers found they could eliminate hollow defects and recover small lesions that standard models often truncate.
  • Efficiency via Prompting: The GroupPrompt3D module allows a model trained for only 200 epochs to outperform a 1000-epoch baseline. This suggests that providing the model with "context" (the support volume) is a massive shortcut for learning.
  • Language as a Regularizer: Using a "many-to-one" mapping for synonyms doesn't just fix naming issues; it actually helps the model learn general visual features. By grouping synonyms, the model sees more "examples" of the same underlying concept, which "smooths" the latent space.

Key Results

  • Zero-Shot Success: On a completely unseen "kidney tumor" category, the Many-to-One CSH achieved a Dice score of 0.139, whereas standard models scored 0. This proves the model can leverage semantic relationships to find lesions it wasn't explicitly trained on.
  • Speed and Accuracy: Using 60 prompts, the model reached competitive Dice values in just 21 hours of training, compared to the 104 hours required for the 1000-epoch baseline.
  • Robustness: The CSH model maintained stable performance even when category names were replaced with complex clinical synonyms or intentional typos (e.g., "ancreatic malignant neoplasm").

Limitations and Future Work

While these modules provide a significant boost, the authors acknowledge that highly irregular tiny lesions still pose a challenge. The zero-shot performance, while a breakthrough, is not yet high enough for "high-risk clinical diagnosis" or automated surgical planning without human oversight.

The authors suggest that the next step is progressive label-wise incremental learning—a two-stage process where the model first adapts to a few samples of a new category before undergoing a full training stage to stabilize its knowledge.

Summary updated

Abstract

This work introduces three lightweight enhancement modules for medical lesion segmentation to improve clinical applicability, particularly under data-scarce and computationally constrained environments. Current mainstream methods struggle with the lack of high-quality annotations for rare pathologies, which often leads to severe morphological distortions during inference, despite achieving seemingly ideal DSC scores. Furthermore, the isolated development of individual segmentation tasks restricts cross-domain methodological innovation, underscoring the need for leveraging shared knowledge across different lesions within the same anatomical region. However, constrained access to historical data and annotations hinders efficient knowledge extraction and triggers catastrophic forgetting when adapting to new classes. To address these challenges, we first propose DT-Morph Combined Loss with continuous spatial pooling to optimize differentiable topological structures and boundary continuity. Second, we introduce 3D Group Prompt, which extracts domain-universal knowledge via curated volumetric image-mask alignment prompts together with text prompts built on SAT baseline. Finally, we replace conventional output layers with Many-to-One Conditional Segmentation Head equipped with synonym-resolved routing, which brings negligible computational cost and enables fast incremental learning for new categories without sacrificing performance on old tasks. Ablation studies on abdominal and extra-abdominal datasets covering both organ and lesion segmentation verify the efficacy of our modules.
BibTeX
@misc{li2026enhancing,
  author = {Ruichen Li and Zhao Yao},
  title = {{Enhancing Medical Lesion Segmentation via Several Simple Modules}},
  year = {2026},
  eprint = {2610.41652},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.41652},
}
Figure