The Learnability Gap in Medical Latent Diffusion

• URL: http://arxiv.org/abs/2605.17087v1
• Date: Sat, 16 May 2026 17:07:54 GMT
• Title: The Learnability Gap in Medical Latent Diffusion
• Abstract summary: Generative data augmentation with latent diffusion models is a promising strategy for addressing class imbalance in medical imaging.<n>We identify and formalize the learnability gap: large-scale pretrained autoencoders faithfully encode discriminative features for medical classification.<n>We develop noise-conditioned latent classifiers with FiLM layers and image-space distillation that offer 64x throughput and 120x memory gains over image-space models.
• Score: 10.504309161945065
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Generative data augmentation with latent diffusion models is a promising strategy for addressing class imbalance in medical imaging, yet current approaches focus on perceptual fidelity and domain-specific autoencoder fine-tuning while neglecting a more fundamental bottleneck. We identify and formalize the learnability gap: large-scale pretrained autoencoders faithfully encode discriminative features for medical classification, as evidenced by near-lossless performance in reconstruction space, yet their latent representations are structured in ways that are difficult for classifiers to learn from. Across five autoencoder families and four medical benchmarks spanning chest radiography, dermatoscopy, computed tomography, and echocardiography, we show that this gap persists regardless of architecture, initialization strategy, or hyperparameter tuning, and that medical-domain fine-tuning of the autoencoder does not close it. To probe and partially narrow the gap, we develop noise-conditioned latent classifiers with FiLM layers and image-space distillation that offer 64x throughput and 120x memory gains over image-space models while serving as diagnostic tools for latent space quality. Our analysis provides a new framework for evaluating autoencoder latent spaces and identifies their structure, rather than their fidelity or domain specificity, as the primary obstacle to closing the performance gap between real and synthetic medical training data.

Related papers

• Representation-driven Endoscopic Visual Embedding Alignment for Latent Generation [9.107208576219668]
  We introduce REVEAL, the largest generative foundation model for endoscopy to date.<n>RevEAL employs encoders pretrained directly on the endoscopic distribution to align diffusion latents with domain-specific visual features.<n>RevEAL produces high-fidelity images and maintains robust structural coherence in latent-space edits.
  arXiv  Detail & Related papers  (2026-08-07T12:47:36Z)
• Dementia Etiology Diagnosis via Collaborative Meta Knowledge Enhancement [81.27196033392559]
  We propose a Collaborative Meta Knowledge Enhancement (COME) framework for dementia etiology diagnosis.<n>COME injects multi-center acquisition semantics, source identifiers, and modality indicators as heterogeneity-aware embeddings into a unified Transformer architecture for scale-up training.<n>Our method achieves state-of-the-art in-domain performance with a mean macro-averaged AUC of 85.62% and a 4.29-point gain over the strongest baseline.
  arXiv  Detail & Related papers  (2026-07-24T04:26:47Z)
• CIPHER: Causal Intervention Pathways for Healthcare Equity and Robustness [12.8473199678431]
  We formalize the medical image formation process via a structural causal model.<n>We introduce CIPHER, a framework designed to intervene on all four causal paths.<n>By employing a multi-pathway intervention strategy, our model reduced worst-group disparities by an average of 35.8%.
  arXiv  Detail & Related papers  (2026-07-01T09:36:53Z)
• Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay [29.68388420763425]
  This paper presents the first continual learning framework specifically designed for fMRI-based diagnosis across heterogeneous clinical sites.<n>Our framework introduces a structure-aware variational autoencoder that synthesizes realistic FC matrices for both patient and control groups.<n> Experiments on multi-site datasets for major depressive disorder (MDD), schizophrenia (SZ), and autism spectrum disorder (ASD) show that the proposed generative model enhances data augmentation quality.
  arXiv  Detail & Related papers  (2026-04-15T16:08:52Z)
• Delving Aleatoric Uncertainty in Medical Image Segmentation via Vision Foundation Models [56.29123284262618]
  This work proposes leveraging the universal representation capabilities of visual foundation models to estimate inherent data uncertainty.<n>We analyze the feature diversity of the model's decoded representations and quantify their singular value energy to define the semantic perception scale for each class.<n>Based on this foundation, we design two uncertainty-driven application strategies: (1) the aleatoric uncertainty-aware data filtering mechanism to eliminate potentially noisy samples and enhance model learning quality; and (2) the dynamic uncertainty-aware optimization strategy that adaptively adjusts class-specific loss weights during training based on the semantic perception scale, combined with a label denoising mechanism to improve training stability
  arXiv  Detail & Related papers  (2026-04-13T03:59:54Z)
• Adaptive Clinical-Aware Latent Diffusion for Multimodal Brain Image Generation and Missing Modality Imputation [12.941925130242039]
  ACADiff is a framework that synthesizes missing brain imaging modalities.<n>It learns mappings between incomplete multimodal observations and target modalities by progressively denoising latent representations.<n>It achieves superior generation quality and maintains robust diagnostic performance even under extreme 80% missing scenarios.
  arXiv  Detail & Related papers  (2026-03-10T17:26:45Z)
• Deep Expert Injection for Anchoring Retinal VLMs with Domain-Specific Knowledge [66.67024684187915]
  Large Vision Language Models (LVLMs) show immense potential for automated ophthalmic diagnosis.<n>Their clinical deployment is severely hindered by lacking domain-specific knowledge.<n>EyExIn is a framework designed to anchor retinal VLMs with expert knowledge via a Deep Expert Injection mechanism.
  arXiv  Detail & Related papers  (2026-03-07T09:43:49Z)
• A Semantically Enhanced Generative Foundation Model Improves Pathological Image Synthesis [82.01597026329158]
  We introduce a Correlation-Regulated Alignment Framework for Tissue Synthesis (CRAFTS) for pathology-specific text-to-image synthesis.<n>CRAFTS incorporates a novel alignment mechanism that suppresses semantic drift to ensure biological accuracy.<n>This model generates diverse pathological images spanning 30 cancer types, with quality rigorously validated by objective metrics and pathologist evaluations.
  arXiv  Detail & Related papers  (2025-12-15T10:22:43Z)
• Robust Noisy Pseudo-label Learning for Semi-supervised Medical Image Segmentation Using Diffusion Model [5.158113225132093]
  Semi-supervised medical image segmentation aims to leverage limited annotated data alongside abundant unlabeled data to achieve accurate segmentation.<n>Existing methods often struggle to structure semantic distributions in the latent space due to noise introduced by pseudo-labels.<n>Our method introduces a constraint into the latent structure of semantic labels during the denoising diffusion process by enforcing prototype-based contrastive consistency.
  arXiv  Detail & Related papers  (2025-07-22T10:21:55Z)
• Self-supervised Learning of Echocardiographic Video Representations via Online Cluster Distillation [21.738308923180767]
  We present DISCOVR, a self-supervised dual branch framework for cardiac ultrasound video representation learning.<n>DISCOVR combines a clustering-based video encoder that models temporal dynamics with an online image encoder that extracts fine-grained spatial semantics.
  arXiv  Detail & Related papers  (2025-06-13T13:36:33Z)
• Multi-Scale Feature Fusion with Image-Driven Spatial Integration for Left Atrium Segmentation from Cardiac MRI Images [0.0]
  We propose a framework that integrates DINOv2 as an encoder with a UNet-style decoder, incorporating multi-scale feature fusion and input image integration.<n>We validate our approach on the LAScarQS 2022 dataset and demonstrate improved performance with a 92.3% Dice and 84.1% IoU score for giant architecture.
  arXiv  Detail & Related papers  (2025-02-10T16:12:46Z)
• CathFlow: Self-Supervised Segmentation of Catheters in Interventional Ultrasound Using Optical Flow and Transformers [66.15847237150909]
  We introduce a self-supervised deep learning architecture to segment catheters in longitudinal ultrasound images. The network architecture builds upon AiAReSeg, a segmentation transformer built with the Attention in Attention mechanism. We validated our model on a test dataset, consisting of unseen synthetic data and images collected from silicon aorta phantoms.
  arXiv  Detail & Related papers  (2024-03-21T15:13:36Z)
• G-MIND: An End-to-End Multimodal Imaging-Genetics Framework for Biomarker Identification and Disease Classification [49.53651166356737]
  We propose a novel deep neural network architecture to integrate imaging and genetics data, as guided by diagnosis, that provides interpretable biomarkers. We have evaluated our model on a population study of schizophrenia that includes two functional MRI (fMRI) paradigms and Single Nucleotide Polymorphism (SNP) data.
  arXiv  Detail & Related papers  (2021-01-27T19:28:04Z)

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.