EYE-Llama, an in-domain large language model for ophthalmology

Summary: Training large language models (LLMs) on domain-specific data enhances their performance, yielding more accurate and reliable question-answering (Q&A) systems that support clinical decision-making and patient education. We present EYE-Llama, pretrained on ophthalmology-focused datasets,...

Full description

Saved in:
Bibliographic Details
Main Authors: Tania Haghighi, Sina Gholami, Jared Todd Sokol, Enaika Kishnani, Adnan Ahsaniyan, Holakou Rahmanian, Fares Hedayati, Theodore Leng, Minhaj Nur Alam
Format: Article
Language:English
Published: Elsevier 2025-07-01
Series:iScience
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2589004225012453
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849318372753276928
author Tania Haghighi
Sina Gholami
Jared Todd Sokol
Enaika Kishnani
Adnan Ahsaniyan
Holakou Rahmanian
Fares Hedayati
Theodore Leng
Minhaj Nur Alam
author_facet Tania Haghighi
Sina Gholami
Jared Todd Sokol
Enaika Kishnani
Adnan Ahsaniyan
Holakou Rahmanian
Fares Hedayati
Theodore Leng
Minhaj Nur Alam
author_sort Tania Haghighi
collection DOAJ
description Summary: Training large language models (LLMs) on domain-specific data enhances their performance, yielding more accurate and reliable question-answering (Q&A) systems that support clinical decision-making and patient education. We present EYE-Llama, pretrained on ophthalmology-focused datasets, including PubMed abstracts, textbooks, and online articles, and fine-tuned on diverse Q&A pairs. We evaluated EYE-Llama against Llama 2, Llama 3, Meditron, ChatDoctor, ChatGPT, and several other LLMs. Using BERT (Bidirectional Encoder Representations from Transformers) score, BART (Bidirectional and Auto-Regressive Transformer) score, and BLEU (Bilingual Evaluation Understudy) metrics, EYE-Llama achieved superior scores. On the MedMCQA benchmark, it outperformed Llama 2, Meditron, and ChatDoctor. On PubMedQA, it achieved 0.96 accuracy, surpassing all models tested. These results demonstrate that domain-specific pretraining and fine-tuning significantly improve medical Q&A performance and underscore the value of specialized models such as EYE-Llama.
format Article
id doaj-art-3bddd21bc98b4973b5ee722d601e3f96
institution Kabale University
issn 2589-0042
language English
publishDate 2025-07-01
publisher Elsevier
record_format Article
series iScience
spelling doaj-art-3bddd21bc98b4973b5ee722d601e3f962025-08-20T03:50:50ZengElsevieriScience2589-00422025-07-0128711298410.1016/j.isci.2025.112984EYE-Llama, an in-domain large language model for ophthalmologyTania Haghighi0Sina Gholami1Jared Todd Sokol2Enaika Kishnani3Adnan Ahsaniyan4Holakou Rahmanian5Fares Hedayati6Theodore Leng7Minhaj Nur Alam8Department of Electrical and Computer Engineering, University of North Carolina at Charlotte, Charlotte, NC, USA; Department of Computer Science, Baha’i Institute for Higher Education, Tehran, IranDepartment of Electrical and Computer Engineering, University of North Carolina at Charlotte, Charlotte, NC, USADepartment of Ophthalmology Stanford University School of Medicine, Stanford, CA, USADepartment of Electrical and Computer Engineering, University of North Carolina at Charlotte, Charlotte, NC, USADepartment of Computer Science, Baha’i Institute for Higher Education, Tehran, IranDepartment of Computer Science, Baha’i Institute for Higher Education, Tehran, IranDepartment of Computer Science, Baha’i Institute for Higher Education, Tehran, IranDepartment of Ophthalmology Stanford University School of Medicine, Stanford, CA, USADepartment of Electrical and Computer Engineering, University of North Carolina at Charlotte, Charlotte, NC, USA; Corresponding authorSummary: Training large language models (LLMs) on domain-specific data enhances their performance, yielding more accurate and reliable question-answering (Q&A) systems that support clinical decision-making and patient education. We present EYE-Llama, pretrained on ophthalmology-focused datasets, including PubMed abstracts, textbooks, and online articles, and fine-tuned on diverse Q&A pairs. We evaluated EYE-Llama against Llama 2, Llama 3, Meditron, ChatDoctor, ChatGPT, and several other LLMs. Using BERT (Bidirectional Encoder Representations from Transformers) score, BART (Bidirectional and Auto-Regressive Transformer) score, and BLEU (Bilingual Evaluation Understudy) metrics, EYE-Llama achieved superior scores. On the MedMCQA benchmark, it outperformed Llama 2, Meditron, and ChatDoctor. On PubMedQA, it achieved 0.96 accuracy, surpassing all models tested. These results demonstrate that domain-specific pretraining and fine-tuning significantly improve medical Q&A performance and underscore the value of specialized models such as EYE-Llama.http://www.sciencedirect.com/science/article/pii/S2589004225012453OphthalmologyArtificial intelligence
spellingShingle Tania Haghighi
Sina Gholami
Jared Todd Sokol
Enaika Kishnani
Adnan Ahsaniyan
Holakou Rahmanian
Fares Hedayati
Theodore Leng
Minhaj Nur Alam
EYE-Llama, an in-domain large language model for ophthalmology
iScience
Ophthalmology
Artificial intelligence
title EYE-Llama, an in-domain large language model for ophthalmology
title_full EYE-Llama, an in-domain large language model for ophthalmology
title_fullStr EYE-Llama, an in-domain large language model for ophthalmology
title_full_unstemmed EYE-Llama, an in-domain large language model for ophthalmology
title_short EYE-Llama, an in-domain large language model for ophthalmology
title_sort eye llama an in domain large language model for ophthalmology
topic Ophthalmology
Artificial intelligence
url http://www.sciencedirect.com/science/article/pii/S2589004225012453
work_keys_str_mv AT taniahaghighi eyellamaanindomainlargelanguagemodelforophthalmology
AT sinagholami eyellamaanindomainlargelanguagemodelforophthalmology
AT jaredtoddsokol eyellamaanindomainlargelanguagemodelforophthalmology
AT enaikakishnani eyellamaanindomainlargelanguagemodelforophthalmology
AT adnanahsaniyan eyellamaanindomainlargelanguagemodelforophthalmology
AT holakourahmanian eyellamaanindomainlargelanguagemodelforophthalmology
AT fareshedayati eyellamaanindomainlargelanguagemodelforophthalmology
AT theodoreleng eyellamaanindomainlargelanguagemodelforophthalmology
AT minhajnuralam eyellamaanindomainlargelanguagemodelforophthalmology