Partnerzy

Streszczenie:
Biomedical image datasets are often scarce, expensive to annotate, and vary in
quality due to differences in imaging hardware and techniques. Generative models, particularly diffusion models, have recently demonstrated strong potential to
synthesize realistic medical images, offering a promising strategy for data augmentation. Yet, their application in clinical contexts requires careful validation, as trust,
interpretability, and reliability are essential when medical decisions are at stake.
This work introduces a human-in-the-loop framework for assessing the reliability
and risks of diffusion models in generating breast ultrasound cancer images. Using a Denoising Diffusion Probabilistic Model (D-DDPM), we jointly generate
ultrasound images and corresponding tumor masks from two benchmark datasets
(BUS-BRA and UDIAT). The evaluation pipeline integrates quantitative image
quality metrics (FID, IS, KID), radiologist interpretation, inter-rater agreement
(Cohen’s/Fleiss’ Kappa, Krippendorff’s Alpha), and alignment with large language
model (LLM) outputs. Results show that while D-DDPM can produce images
that are visually similar to real data and sometimes yield higher agreement among
experts than original images, inter-rater reliability remains weak, particularly for
malignant tumors. Radiologists consistently outperform LLMs in classification,
though majority voting across experts improves diagnostic accuracy. These findings highlight both the promise and risks of diffusion models in medical imaging,
including that synthetic ultrasound data can supplement limited datasets; however,
robust expert validation remains indispensable to ensure clinical trustworthiness
and safe integration.