Adapting foundation models for medical image generation
Abstract
Recent advances in vision–language foundation models have enabled new possibilities in high-resolution, semantically aligned medical image synthesis, an increasingly important capability for improving data efficiency, robustness, and transparency in clinical AI systems. Yet, existing approaches remain limited by three key challenges: (1) the absence of principled methods for disentangling and controlling latent factors of variation in high-resolution medical images using vision-language foundation models. (2) the absence of sufficiently high-resolution medical image generation models hinders the faithful synthesis of clinically relevant fine-grained anatomical details, limiting their suitability for diagnostic scenarios that depend on subtle structural cues. (3) the large number of available fine-tuning paradigms makes it difficult to choose an appropriate strategy, and there is currently no comprehensive overview of how these methods affect downstream performance across different clinical contexts. This thesis addresses these gaps through three complementary contributions. First, I propose a novel framework for language-guided latent trajectory traversal in fine-tuned Stable Diffusion models. I demonstrate that these models learn disentangled directions corresponding to medical attributes such as pleural effusion, support devices, and dermatological artifacts. To quantify disentanglement, I introduce the Classifier Flip Rate along a Trajectory (CFRT) and show that latent…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.