DAM-Seg: Anatomically accurate cardiac segmentation using Dense Associative Networks
Zahid Ullah, Jihie Kim
TL;DR
This work addresses anatomically incorrect cardiac segmentations by introducing DAM-Seg, a transformer-based segmentation model augmented with dense associative memory that stores a small set of general cardiac patterns and retrieves them via $m$ meta-stable states to enforce anatomical plausibility. The static memory $oldsymbol{\xi}$ generates keys and values ($K = W_k \boldsymbol{\xi}$, $V = W_q \boldsymbol{\xi}$) and, combined with input queries $Q$, yields $z = V \text{Softmax}((Q^T K)/\sqrt{d})$, guiding the network toward anatomically consistent outputs. The authors provide learnable memory transformations, promote $m$ meta-stable states, and integrate the memory with the DPT architecture, achieving robust improvements on CAMUS and CardiacNet, with ablation and SOTA comparisons supporting the approach. The method enhances segmentation reliability in scenarios with partial visibility and occlusion, offering a practical path toward more dependable cardiac imaging tools, and the authors share data/code availability to support reproducibility.
Abstract
Deep learning-based cardiac segmentation has seen significant advancements over the years. Many studies have tackled the challenge of anatomically incorrect segmentation predictions by introducing auxiliary modules. These modules either post-process segmentation outputs or enforce consistency between specific points to ensure anatomical correctness. However, such approaches often increase network complexity, require separate training for these modules, and may lack robustness in scenarios with poor visibility. To address these limitations, we propose a novel transformer-based architecture that leverages dense associative networks to learn and retain specific patterns inherent to cardiac inputs. Unlike traditional methods, our approach restricts the network to memorize a limited set of patterns. During forward propagation, a weighted sum of these patterns is used to enforce anatomical correctness in the output. Since these patterns are input-independent, the model demonstrates enhanced robustness, even in cases with poor visibility. The proposed pipeline was evaluated on two publicly available datasets, CAMUS and CardiacNet. Experimental results indicate that our model consistently outperforms baseline approaches across all metrics, highlighting its effectiveness and reliability for cardiac segmentation tasks.