A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs
Abstract Rare diseases, affecting ~350 million people worldwide, pose significant challenges in clinical diagnosis due to the lack of experienced physicians and the complexity of differentiating between numerous rare diseases. To address these challenges, we introduce PhenoBrain, a fully automated a...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | npj Digital Medicine |
| Online Access: | https://doi.org/10.1038/s41746-025-01452-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832571316160430080 |
|---|---|
| author | Xiaohao Mao Yu Huang Ye Jin Lun Wang Xuanzhong Chen Honghong Liu Xinglin Yang Haopeng Xu Xiaodong Luan Ying Xiao Siqin Feng Jiahao Zhu Xuegong Zhang Rui Jiang Shuyang Zhang Ting Chen |
| author_facet | Xiaohao Mao Yu Huang Ye Jin Lun Wang Xuanzhong Chen Honghong Liu Xinglin Yang Haopeng Xu Xiaodong Luan Ying Xiao Siqin Feng Jiahao Zhu Xuegong Zhang Rui Jiang Shuyang Zhang Ting Chen |
| author_sort | Xiaohao Mao |
| collection | DOAJ |
| description | Abstract Rare diseases, affecting ~350 million people worldwide, pose significant challenges in clinical diagnosis due to the lack of experienced physicians and the complexity of differentiating between numerous rare diseases. To address these challenges, we introduce PhenoBrain, a fully automated artificial intelligence pipeline. PhenoBrain utilizes a BERT-based natural language processing model to extract phenotypes from clinical texts in EHRs and employs five new diagnostic models for differential diagnoses of rare diseases. The AI system was developed and evaluated on diverse, multi-country rare disease datasets, comprising 2271 cases with 431 rare diseases. In 1936 test cases, PhenoBrain achieved an average predicted top-3 recall of 0.513 and a top-10 recall of 0.654, surpassing 13 leading prediction methods. In a human-computer study with 75 cases, PhenoBrain exhibited exceptional performance with a top-3 recall of 0.613 and a top-10 recall of 0.813, surpassing the performance of 50 specialist physicians and large language models like ChatGPT and GPT-4. Combining PhenoBrain’s predictions with specialists increased the top-3 recall to 0.768, demonstrating its potential to enhance diagnostic accuracy in clinical workflows. |
| format | Article |
| id | doaj-art-928ab9111ce24cde9c384c5d71a14586 |
| institution | Kabale University |
| issn | 2398-6352 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Digital Medicine |
| spelling | doaj-art-928ab9111ce24cde9c384c5d71a145862025-02-02T12:43:41ZengNature Portfolionpj Digital Medicine2398-63522025-01-018111510.1038/s41746-025-01452-1A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRsXiaohao Mao0Yu Huang1Ye Jin2Lun Wang3Xuanzhong Chen4Honghong Liu5Xinglin Yang6Haopeng Xu7Xiaodong Luan8Ying Xiao9Siqin Feng10Jiahao Zhu11Xuegong Zhang12Rui Jiang13Shuyang Zhang14Ting Chen15Department of Computer Science and Technology & Institute for Artificial Intelligence & BNRist, Tsinghua UniversityDepartment of Computer Science and Technology & Institute for Artificial Intelligence & BNRist, Tsinghua UniversityMedical Research Center, State Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Internal Medicine, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Computer Science and Technology & Institute for Artificial Intelligence & BNRist, Tsinghua UniversityDepartment of Internal Medicine, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Cardiology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeNuffield Department of Medicine, University of OxfordState Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Geriatrics, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Cardiology, State Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Computer Science and Technology & Institute for Artificial Intelligence & BNRist, Tsinghua UniversityDepartment of Automation & BNRist, Tsinghua UniversityDepartment of Automation & BNRist, Tsinghua UniversityDepartment of Cardiology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical CollegeDepartment of Computer Science and Technology & Institute for Artificial Intelligence & BNRist, Tsinghua UniversityAbstract Rare diseases, affecting ~350 million people worldwide, pose significant challenges in clinical diagnosis due to the lack of experienced physicians and the complexity of differentiating between numerous rare diseases. To address these challenges, we introduce PhenoBrain, a fully automated artificial intelligence pipeline. PhenoBrain utilizes a BERT-based natural language processing model to extract phenotypes from clinical texts in EHRs and employs five new diagnostic models for differential diagnoses of rare diseases. The AI system was developed and evaluated on diverse, multi-country rare disease datasets, comprising 2271 cases with 431 rare diseases. In 1936 test cases, PhenoBrain achieved an average predicted top-3 recall of 0.513 and a top-10 recall of 0.654, surpassing 13 leading prediction methods. In a human-computer study with 75 cases, PhenoBrain exhibited exceptional performance with a top-3 recall of 0.613 and a top-10 recall of 0.813, surpassing the performance of 50 specialist physicians and large language models like ChatGPT and GPT-4. Combining PhenoBrain’s predictions with specialists increased the top-3 recall to 0.768, demonstrating its potential to enhance diagnostic accuracy in clinical workflows.https://doi.org/10.1038/s41746-025-01452-1 |
| spellingShingle | Xiaohao Mao Yu Huang Ye Jin Lun Wang Xuanzhong Chen Honghong Liu Xinglin Yang Haopeng Xu Xiaodong Luan Ying Xiao Siqin Feng Jiahao Zhu Xuegong Zhang Rui Jiang Shuyang Zhang Ting Chen A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs npj Digital Medicine |
| title | A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs |
| title_full | A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs |
| title_fullStr | A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs |
| title_full_unstemmed | A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs |
| title_short | A phenotype-based AI pipeline outperforms human experts in differentially diagnosing rare diseases using EHRs |
| title_sort | phenotype based ai pipeline outperforms human experts in differentially diagnosing rare diseases using ehrs |
| url | https://doi.org/10.1038/s41746-025-01452-1 |
| work_keys_str_mv | AT xiaohaomao aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yuhuang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yejin aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT lunwang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xuanzhongchen aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT honghongliu aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xinglinyang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT haopengxu aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xiaodongluan aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yingxiao aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT siqinfeng aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT jiahaozhu aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xuegongzhang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT ruijiang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT shuyangzhang aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT tingchen aphenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xiaohaomao phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yuhuang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yejin phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT lunwang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xuanzhongchen phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT honghongliu phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xinglinyang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT haopengxu phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xiaodongluan phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT yingxiao phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT siqinfeng phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT jiahaozhu phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT xuegongzhang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT ruijiang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT shuyangzhang phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs AT tingchen phenotypebasedaipipelineoutperformshumanexpertsindifferentiallydiagnosingrarediseasesusingehrs |