Leveraging foundation models for content-based image retrieval in radiology.

Comput Biol Med

German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Heidelberg, Germany; Faculty of Mathematics and Computer Science, Heidelberg University, Heidelberg, Germany; Medical Faculty Heidelberg, University of Heidelberg, Heidelberg, Germany.

Published: September 2025


Category Ranking

98%

Total Visits

921

Avg Visit Duration

2 minutes

Citations

20

Article Abstract

Content-based image retrieval (CBIR) has the potential to significantly improve diagnostic aid and medical research in radiology. However, current CBIR systems face limitations due to their specialization to certain pathologies, limiting their utility. On the other hand, several vision foundation models have been shown to produce general-purpose visual features. Therefore, in this work, we propose using vision foundation models as powerful and versatile off-the-shelf feature extractors for content-based image retrieval. Our contributions include: (1) benchmarking a diverse set of vision foundation models on an extensive dataset comprising 1.6 million 2D radiological images across four modalities and 161 pathologies; (2) identifying weakly-supervised models, particularly BiomedCLIP, as highly effective, achieving a P@1 of up to 0.594 (P@3: 0.590, P@5: 0.588, P@10: 0.583), comparable to specialized CBIR systems but without additional training; (3) conducting an in-depth analysis of the impact of index size on retrieval performance; (4) evaluating the quality of embedding spaces generated by different models; and (5) investigating specific challenges associated with retrieving anatomical versus pathological structures. Despite these challenges, our research underscores the vast potential of foundation models for CBIR in radiology, proposing a shift towards versatile, general-purpose medical image retrieval systems that do not require specific tuning. Our code, dataset splits and embeddings are publicly available here.

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.compbiomed.2025.110640DOI Listing

Publication Analysis

Top Keywords

foundation models
20
image retrieval
16
content-based image
12
vision foundation
12
cbir systems
8
models
7
retrieval
5
leveraging foundation
4
models content-based
4
image
4

Similar Publications

CRISPR/Cas9-mediated editing of COQ4 in induced pluripotent stem cells: A model for investigating COQ4-associated human coenzyme Q deficiency.

Stem Cell Res

September 2025

Department of General Pediatrics, Neonatology, and Pediatric Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University, Düsseldorf 40225, Germany. Electronic address:

Pathogenic variants in the gene COQ4 cause primary coenzyme Q deficiency, which is associated with symptoms ranging from early epileptic encephalopathy up to adult-onset ataxia-spasticity spectrum disease. We genetically modified commercially available wild-type iPS cells by using a CRISPR/Cas9 approach to create heterozygous and homozygous isogenic cell lines carrying the disease-causing COQ4 variants c.458C > T, p.

View Article and Find Full Text PDF

Rationale: There are insufficient data to inform the management of central sleep apnea (CSA) in patients with heart failure (HF) with reduced ejection fraction (HFrEF). Nocturnal oxygen therapy (NOT) has been postulated to benefit CSA patients with HFrEF, but has not been rigorously studied. This article is open access and distributed under the terms of the Creative Commons Attribution Non-Commercial No Derivatives License 4.

View Article and Find Full Text PDF

Predicting Unplanned Readmission Risk in Patients With Cirrhosis: Complication-Aware Dynamic Classifier Selection Approach.

JMIR Med Inform

September 2025

College of Medical Informatics, Chongqing Medical University, 1 Yixueyuan Road, Yuzhong District, Chongqing, 400016, China, 86 13500303273.

Background: Cirrhosis is a leading cause of noncancer deaths in gastrointestinal diseases, resulting in high hospitalization and readmission rates. Early identification of high-risk patients is vital for proactive interventions and improving health care outcomes. However, the quality and integrity of real-world electronic health records (EHRs) limit their utility in developing risk assessment tools.

View Article and Find Full Text PDF

Objective: To clarify the potential risks and causative mechanisms of glare from nighttime road fill lights on driving safety, this study investigates the dual interference of glare-induced visual cognitive load and physiological stress.

Methods: A field driving experiment involving 20 drivers was conducted, with real-time collection of visual data (e.g.

View Article and Find Full Text PDF

Physical activity and salutogenesis: a necessary dialogue.

Cien Saude Colet

August 2025

Departamento de Educação Física, Núcleo de Saúde, Universidade Federal de Rondônia. Porto Velho RO Brasil.

This essay aims to broaden the dialogue between physical activity (PA) and health by a salutogenesis-oriented concept, which emphasizes the origins of health, as opposed to the pathogenic model, which focuses on the origin of diseases. It presents the foundations of the theory, created by Aaron Antonovsky, whose health promotion model is centered on the sense of coherence (SOC), which consists of significance, comprehensibility, and manageability. Then, it explores the relationships between PA and salutogenesis, demonstrating how this framework can be applied in various life stages and practice settings.

View Article and Find Full Text PDF