External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets
Retinal hemorrhage (RH) is a significant clinical finding with various etiologies, necessitating accurate classification for effective management. This study aims to externally validate deep learning (DL) models, specifically FastVit_SA12 and ResNet18, for distinguishing between traumatic and medica...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-12-01
|
| Series: | Bioengineering |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2306-5354/12/1/20 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832589078068985856 |
|---|---|
| author | Pooya Khosravi Nolan A. Huck Kourosh Shahraki Elina Ghafari Reza Azimi So Young Kim Eric Crouch Xiaohui Xie Donny W. Suh |
| author_facet | Pooya Khosravi Nolan A. Huck Kourosh Shahraki Elina Ghafari Reza Azimi So Young Kim Eric Crouch Xiaohui Xie Donny W. Suh |
| author_sort | Pooya Khosravi |
| collection | DOAJ |
| description | Retinal hemorrhage (RH) is a significant clinical finding with various etiologies, necessitating accurate classification for effective management. This study aims to externally validate deep learning (DL) models, specifically FastVit_SA12 and ResNet18, for distinguishing between traumatic and medical causes of RH using diverse fundus photography datasets. A comprehensive dataset was compiled, including private collections from South Korea and Virginia, alongside publicly available datasets such as RFMiD, BRSET, and DeepEyeNet. The models were evaluated on a total of 2661 images, achieving high performance metrics. FastVit_SA12 demonstrated an overall accuracy of 96.99%, with a precision of 0.9935 and recall of 0.9723 for medical cases, while ResNet18 achieved a 94.66% accuracy with a precision of 0.9893. A Grad-CAM analysis revealed that ResNet18 emphasized global vascular patterns, such as arcuate vessels, while FastVit_SA12 focused on clinically relevant areas, including the optic disk and hemorrhagic regions. Medical cases showed localized activations, whereas trauma-related images displayed diffuse patterns across the fundus. Both models exhibited strong sensitivity and specificity, indicating their potential utility in clinical settings for accurate RH diagnosis. This study underscores the importance of external validation in enhancing the reliability and applicability of AI models in ophthalmology, paving the way for improved patient care and outcomes. |
| format | Article |
| id | doaj-art-d21fa339dabb4299abad6af1f25520e7 |
| institution | Kabale University |
| issn | 2306-5354 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Bioengineering |
| spelling | doaj-art-d21fa339dabb4299abad6af1f25520e72025-01-24T13:22:59ZengMDPI AGBioengineering2306-53542024-12-011212010.3390/bioengineering12010020External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography DatasetsPooya Khosravi0Nolan A. Huck1Kourosh Shahraki2Elina Ghafari3Reza Azimi4So Young Kim5Eric Crouch6Xiaohui Xie7Donny W. Suh8Department of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, Soonchunhyang University College of Medicine, Cheonan, 31151, Republic of KoreaDepartment of Ophthalmology, Eastern Virginia Medical School, Norfolk, VA 23507, USADonald Bren School of Information and Computer Sciences, University of California, Irvine, CA 92697, USADepartment of Ophthalmology, School of Medicine, University of California, Irvine, CA 92697, USARetinal hemorrhage (RH) is a significant clinical finding with various etiologies, necessitating accurate classification for effective management. This study aims to externally validate deep learning (DL) models, specifically FastVit_SA12 and ResNet18, for distinguishing between traumatic and medical causes of RH using diverse fundus photography datasets. A comprehensive dataset was compiled, including private collections from South Korea and Virginia, alongside publicly available datasets such as RFMiD, BRSET, and DeepEyeNet. The models were evaluated on a total of 2661 images, achieving high performance metrics. FastVit_SA12 demonstrated an overall accuracy of 96.99%, with a precision of 0.9935 and recall of 0.9723 for medical cases, while ResNet18 achieved a 94.66% accuracy with a precision of 0.9893. A Grad-CAM analysis revealed that ResNet18 emphasized global vascular patterns, such as arcuate vessels, while FastVit_SA12 focused on clinically relevant areas, including the optic disk and hemorrhagic regions. Medical cases showed localized activations, whereas trauma-related images displayed diffuse patterns across the fundus. Both models exhibited strong sensitivity and specificity, indicating their potential utility in clinical settings for accurate RH diagnosis. This study underscores the importance of external validation in enhancing the reliability and applicability of AI models in ophthalmology, paving the way for improved patient care and outcomes.https://www.mdpi.com/2306-5354/12/1/20retinal hemorrhage (RH)deep learning (DL)fundus photographyexternal validation |
| spellingShingle | Pooya Khosravi Nolan A. Huck Kourosh Shahraki Elina Ghafari Reza Azimi So Young Kim Eric Crouch Xiaohui Xie Donny W. Suh External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets Bioengineering retinal hemorrhage (RH) deep learning (DL) fundus photography external validation |
| title | External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets |
| title_full | External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets |
| title_fullStr | External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets |
| title_full_unstemmed | External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets |
| title_short | External Validation of Deep Learning Models for Classifying Etiology of Retinal Hemorrhage Using Diverse Fundus Photography Datasets |
| title_sort | external validation of deep learning models for classifying etiology of retinal hemorrhage using diverse fundus photography datasets |
| topic | retinal hemorrhage (RH) deep learning (DL) fundus photography external validation |
| url | https://www.mdpi.com/2306-5354/12/1/20 |
| work_keys_str_mv | AT pooyakhosravi externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT nolanahuck externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT kouroshshahraki externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT elinaghafari externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT rezaazimi externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT soyoungkim externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT ericcrouch externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT xiaohuixie externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets AT donnywsuh externalvalidationofdeeplearningmodelsforclassifyingetiologyofretinalhemorrhageusingdiversefundusphotographydatasets |