URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis
Summary: Ultrasound imaging is critical for clinical diagnostics, providing insights into various diseases and organs. However, artificial intelligence (AI) in this field faces challenges, such as the need for large labeled datasets and limited task-specific model applicability, particularly due to...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-08-01
|
| Series: | iScience |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2589004225011782 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849713432861868032 |
|---|---|
| author | Qingbo Kang Qicheng Lao Jun Gao Wuyongga Bao Zhu He Chenlin Du Qiang Lu Kang Li |
| author_facet | Qingbo Kang Qicheng Lao Jun Gao Wuyongga Bao Zhu He Chenlin Du Qiang Lu Kang Li |
| author_sort | Qingbo Kang |
| collection | DOAJ |
| description | Summary: Ultrasound imaging is critical for clinical diagnostics, providing insights into various diseases and organs. However, artificial intelligence (AI) in this field faces challenges, such as the need for large labeled datasets and limited task-specific model applicability, particularly due to ultrasound’s low signal-to-noise ratio (SNR). To overcome these, we introduce the Ultrasound Representation Foundation Model (URFM), designed to learn robust, generalizable representations from unlabeled ultrasound images, enabling label-efficient adaptation to diverse diagnostic tasks. URFM is pre-trained on over 1M images from 15 major anatomical organs using representation-based masked image modeling (MIM), an advanced self-supervised learning. Unlike traditional pixel-based MIM, URFM integrates high-level representations from BiomedCLIP, a specialized medical vision-language model, to address the low SNR issue. Extensive evaluation shows that URFM outperforms state-of-the-art methods, offering enhanced generalization, label efficiency, and training-time efficiency. URFM’s scalability and flexibility signal a significant advancement in diagnostic accuracy and clinical workflow optimization in ultrasound imaging. |
| format | Article |
| id | doaj-art-63777c92e1094ca5ac5d46cd23cbd133 |
| institution | DOAJ |
| issn | 2589-0042 |
| language | English |
| publishDate | 2025-08-01 |
| publisher | Elsevier |
| record_format | Article |
| series | iScience |
| spelling | doaj-art-63777c92e1094ca5ac5d46cd23cbd1332025-08-20T03:13:57ZengElsevieriScience2589-00422025-08-0128811291710.1016/j.isci.2025.112917URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosisQingbo Kang0Qicheng Lao1Jun Gao2Wuyongga Bao3Zhu He4Chenlin Du5Qiang Lu6Kang Li7West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu 610041, Sichuan, China; Med-X Center for Informatics, Sichuan University, Chengdu, Sichuan 610041, China; West China Hospital-SenseTime Joint Lab, Chengdu, Sichuan 610041, ChinaSchool of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China; Corresponding authorWest China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu 610041, Sichuan, China; Med-X Center for Informatics, Sichuan University, Chengdu, Sichuan 610041, China; West China Hospital-SenseTime Joint Lab, Chengdu, Sichuan 610041, China; Stork Healthcare, Chengdu 610041, Sichuan, ChinaDepartment of Ultrasonography, West China Hospital, Sichuan University, Chengdu 610041, Sichuan, ChinaSchool of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaSchool of Biomedical Engineering, Tsinghua University, Beijing 100084, ChinaDepartment of Ultrasonography, West China Hospital, Sichuan University, Chengdu 610041, Sichuan, China; Corresponding authorWest China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu 610041, Sichuan, China; Med-X Center for Informatics, Sichuan University, Chengdu, Sichuan 610041, China; West China Hospital-SenseTime Joint Lab, Chengdu, Sichuan 610041, China; Sichuan University - Pittsburgh Institute, Sichuan University, Chengdu, Sichuan 610207, China; Corresponding authorSummary: Ultrasound imaging is critical for clinical diagnostics, providing insights into various diseases and organs. However, artificial intelligence (AI) in this field faces challenges, such as the need for large labeled datasets and limited task-specific model applicability, particularly due to ultrasound’s low signal-to-noise ratio (SNR). To overcome these, we introduce the Ultrasound Representation Foundation Model (URFM), designed to learn robust, generalizable representations from unlabeled ultrasound images, enabling label-efficient adaptation to diverse diagnostic tasks. URFM is pre-trained on over 1M images from 15 major anatomical organs using representation-based masked image modeling (MIM), an advanced self-supervised learning. Unlike traditional pixel-based MIM, URFM integrates high-level representations from BiomedCLIP, a specialized medical vision-language model, to address the low SNR issue. Extensive evaluation shows that URFM outperforms state-of-the-art methods, offering enhanced generalization, label efficiency, and training-time efficiency. URFM’s scalability and flexibility signal a significant advancement in diagnostic accuracy and clinical workflow optimization in ultrasound imaging.http://www.sciencedirect.com/science/article/pii/S2589004225011782Health sciencesComputer-aided diagnosis methodUltrasound technologyComputer scienceArtificial intelligenceMachine learning |
| spellingShingle | Qingbo Kang Qicheng Lao Jun Gao Wuyongga Bao Zhu He Chenlin Du Qiang Lu Kang Li URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis iScience Health sciences Computer-aided diagnosis method Ultrasound technology Computer science Artificial intelligence Machine learning |
| title | URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis |
| title_full | URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis |
| title_fullStr | URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis |
| title_full_unstemmed | URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis |
| title_short | URFM: A general Ultrasound Representation Foundation Model for advancing ultrasound image diagnosis |
| title_sort | urfm a general ultrasound representation foundation model for advancing ultrasound image diagnosis |
| topic | Health sciences Computer-aided diagnosis method Ultrasound technology Computer science Artificial intelligence Machine learning |
| url | http://www.sciencedirect.com/science/article/pii/S2589004225011782 |
| work_keys_str_mv | AT qingbokang urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT qichenglao urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT jungao urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT wuyonggabao urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT zhuhe urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT chenlindu urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT qianglu urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis AT kangli urfmageneralultrasoundrepresentationfoundationmodelforadvancingultrasoundimagediagnosis |