Compositional Reward Models for Conditional Medical Image Generation

• URL: http://arxiv.org/abs/2609.05028v1
• Date: Fri, 04 Sep 2026 11:46:25 GMT
• Title: Compositional Reward Models for Conditional Medical Image Generation
• Abstract summary: We propose PRISM, a Compositional Reward Model for conditional medical image generation.<n>We decompose image quality into verifier grounded stages, each evaluating a distinct aspect of correctness from fine to coarse properties.<n>We evaluate PRISM across three datasets spanning diverse medical imaging tasks.
• Score: 24.272210629538836
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Acquiring high quality annotated medical image data is critical for training deep learning models; however, annotation is expensive, time consuming, and requires domain expertise. Conditional diffusion models, such as ControlNet, offer an alternative by generating images conditioned on semantic masks and text. However, existing approaches fail to capture fine grained properties (e.g., intensity and texture), as well as semantic consistency expected by domain experts, limiting their effectiveness for downstream tasks. Recent attempts to address these issues using reinforcement learning fine-tuning remain limited due to the reliance on a single scalar reward, which conflates diverse failure modes and provides weak corrective signals. We propose PRISM, a Compositional Reward Model (CRM) framework for conditional medical image generation. Instead of assigning a single reward, we decompose image quality into verifier grounded stages, each evaluating a distinct aspect of correctness from fine to coarse properties, including low level attributes (intensity and texture), structural alignment with conditioning inputs, and high level semantic fidelity. These stage wise rewards are composed through a Hierarchical Constrained Propagation (HCP) mechanism that enforces a fine to coarse notion of correctness, ensuring that lower level deficiencies are resolved before higher level rewards are accrued, preventing easier objectives from masking critical failures. We evaluate PRISM across three datasets spanning diverse medical imaging tasks: PanNuke (multi-class cell segmentation), CeDeM (villi/crypt detection and measurement), and ISIC (skin lesion classification). Training downstream models with data generated by PRISM yields improvements over closest baselines, including a 2.3% increase in mDice on PanNuke, a 8.5% reduction in Mean Relative Error (MRE) on CeDeM, and increases ISIC F1 by 5.9%.

Related papers

• CDRRM: Contrast-Driven Rubric Generation for Reliable and Interpretable Reward Modeling [61.75914342638658]
  CDRRM (Contrast-Driven Reward Model) is a framework for high-quality rubric generation and guided preference judgment.<n>This work offers a scalable, interpretable, and data-efficient path for reward modeling.
  arXiv  Detail & Related papers  (2026-03-09T07:15:23Z)
• T3: Test-Time Model Merging in VLMs for Zero-Shot Medical Imaging Analysis [15.624549727053475]
  Existing model-merging techniques fail to deliver consistent gains across diverse medical modalities.<n>We introduce Test-Time Task adaptive merging (T3), a backpropagation-free framework that computes per-sample coefficients.<n>We present a rigorous cross-evaluation protocol spanning in-domain, base-to-novel, and corruptions across four modalities.
  arXiv  Detail & Related papers  (2025-10-31T08:05:40Z)
• A New One-Shot Federated Learning Framework for Medical Imaging Classification with Feature-Guided Rectified Flow and Knowledge Distillation [13.353672721534627]
  One-Shot Federated Learning (OSFL) has attracted increasing attention due to its low communication overhead.<n>Existing generative model-based OSFL methods suffer from low training efficiency and potential privacy leakage in the healthcare domain.<n>In this paper a modified OSFL framework is proposed, in which a new Feature-Guided Rectified Flow Model (FG-RF) and Dual-Layer Knowledge Distillation (DLKD) aggregation method are developed.
  arXiv  Detail & Related papers  (2025-07-25T08:05:47Z)
• Model Inversion Attacks Through Target-Specific Conditional Diffusion Models [54.69008212790426]
  Model inversion attacks (MIAs) aim to reconstruct private images from a target classifier's training set, thereby raising privacy concerns in AI applications. Previous GAN-based MIAs tend to suffer from inferior generative fidelity due to GAN's inherent flaws and biased optimization within latent space. We propose Diffusion-based Model Inversion (Diff-MI) attacks to alleviate these issues.
  arXiv  Detail & Related papers  (2024-07-16T06:38:49Z)
• 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)
• A Practical Framework for Unsupervised Structure Preservation Medical Image Enhancement [9.453554184019108]
  In practice, low-quality (LQ) medical images, such as images that are hazy/blurry, are often obtained during data acquisition. Several generative adversarial networks (GAN)-based image enhancement methods have been proposed and have shown promising results. We propose a framework for practical unsupervised medical image enhancement that includes a non-reference objective evaluation of structure preservation.
  arXiv  Detail & Related papers  (2023-04-04T15:13:44Z)
• Degradation-invariant Enhancement of Fundus Images via Pyramid Constraint Network [27.374391253266428]
  This paper proposes the pyramid constraint to develop a degradation-invariant enhancement network (PCE-Net) High-quality images are randomly degraded to form sequences of low-quality ones sharing the same content (SeqLCs) Then individual low-quality images are decomposed to Laplacian pyramid features (LPF) as the multi-level input for the enhancement.
  arXiv  Detail & Related papers  (2022-10-18T05:45:13Z)
• DOMINO: Domain-aware Model Calibration in Medical Image Segmentation [51.346121016559024]
  Modern deep neural networks are poorly calibrated, compromising trustworthiness and reliability. We propose DOMINO, a domain-aware model calibration method that leverages the semantic confusability and hierarchical similarity between class labels. Our results show that DOMINO-calibrated deep neural networks outperform non-calibrated models and state-of-the-art morphometric methods in head image segmentation.
  arXiv  Detail & Related papers  (2022-09-13T15:31:52Z)
• MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer [62.997667081978825]
  We propose the Multi-instance RST with a drop-max layer, namely MIRST-DM, to learn smoother decision boundaries on small datasets. The proposed approach was validated using a small breast ultrasound dataset with 1,190 images.
  arXiv  Detail & Related papers  (2022-05-02T20:25:26Z)
• Cross-Site Severity Assessment of COVID-19 from CT Images via Domain Adaptation [64.59521853145368]
  Early and accurate severity assessment of Coronavirus disease 2019 (COVID-19) based on computed tomography (CT) images offers a great help to the estimation of intensive care unit event. To augment the labeled data and improve the generalization ability of the classification model, it is necessary to aggregate data from multiple sites. This task faces several challenges including class imbalance between mild and severe infections, domain distribution discrepancy between sites, and presence of heterogeneous features.
  arXiv  Detail & Related papers  (2021-09-08T07:56:51Z)
• Label Geometry Aware Discriminator for Conditional Generative Networks [40.89719383597279]
  Conditional Generative Adversarial Networks (GANs) can generate highly photo realistic images with desired target classes. These synthetic images have not always been helpful to improve downstream supervised tasks such as image classification.
  arXiv  Detail & Related papers  (2021-05-12T08:17:25Z)

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.