Deep Learning–Based Precision Cropping of Eye Regions in Strabismus Photographs: Algorithm Development and Validation Study for Workflow Optimization

Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. This study aimed to address these challenges using an advanced preprocessing al...

Full description

Saved in:
Bibliographic Details
Published in:Journal of medical Internet research Vol. 27; no. 10; p. e74402
Main Authors: Wu, Dawen, Li, Yanfei, Yang, Zeyi, Yin, Teng, Chen, Xiaohang, Liu, Jingyu, Shang, Wenyi, Xie, Bin, Yang, Guoyuan, Zhang, Haixian, Liu, Longqian
Format: Journal Article
Language:English
Published: Canada Journal of Medical Internet Research 17.07.2025
JMIR Publications
Subjects:
ISSN:1438-8871, 1439-4456, 1438-8871
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. This study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing. This retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation. The model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000-1.000), recall of 1.000 (95% CI 1.000-1.000), mAP50 of 0.995 (95% CI 0.995-0.995), and mAP95 of 0.893 (95% CI 0.870-0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901-0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P=.02) and 0.884 (P=.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions. This study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
AbstractList Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application.BackgroundTraditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application.This study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing.ObjectiveThis study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing.This retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation.MethodsThis retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation.The model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000-1.000), recall of 1.000 (95% CI 1.000-1.000), mAP50 of 0.995 (95% CI 0.995-0.995), and mAP95 of 0.893 (95% CI 0.870-0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901-0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P=.02) and 0.884 (P=.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions.ResultsThe model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000-1.000), recall of 1.000 (95% CI 1.000-1.000), mAP50 of 0.995 (95% CI 0.995-0.995), and mAP95 of 0.893 (95% CI 0.870-0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901-0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P=.02) and 0.884 (P=.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions.This study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.ConclusionsThis study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
Background Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. Objective This study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing. Methods This retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation. Results The model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000‐1.000), recall of 1.000 (95% CI 1.000‐1.000), mAP50 of 0.995 (95% CI 0.995‐0.995), and mAP95 of 0.893 (95% CI 0.870‐0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901‐0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P =.02) and 0.884 (P =.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions. Conclusions This study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. This study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing. This retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation. The model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000‐1.000), recall of 1.000 (95% CI 1.000‐1.000), mAP50 of 0.995 (95% CI 0.995‐0.995), and mAP95 of 0.893 (95% CI 0.870‐0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901‐0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P =.02) and 0.884 (P =.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions. This study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. This study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing. This retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation. The model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000-1.000), recall of 1.000 (95% CI 1.000-1.000), mAP50 of 0.995 (95% CI 0.995-0.995), and mAP95 of 0.893 (95% CI 0.870-0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901-0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (P=.02) and 0.884 (P=.05), respectively. Heatmaps highlighted core ocular focus, aligning with head tilt directions. This study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
Abstract BackgroundTraditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial intelligence (AI) model development and clinical application. ObjectiveThis study aimed to address these challenges using an advanced preprocessing algorithm to enhance the accuracy, efficiency, and standardization of eye region cropping for clinical workflows and AI data preprocessing. MethodsThis retrospective and prospective cross-sectional study utilized 5832 images from 648 inpatients and outpatients, capturing 3 gaze positions under diverse conditions, including obstructions and varying distances. The preprocessing algorithm, based on a rotating bounding box detection framework, was trained and evaluated using precision, recall, and mean average precision (mAP) at various intersections over union thresholds. A 5-fold cross-validation was performed on an inpatient dataset, with additional testing on an independent outpatient dataset and an external cross-population dataset of 500 images from the IMDB-WIKI collection, representing diverse ethnicities and ages. Expert validation confirmed alignment with clinical standards across 96 images (48 images from a Chinese dataset of patients with strabismus and 48 images from IMDB-WIKI). Gradient-weighted class activation mapping heatmaps were used to assess model interpretability. A control experiment with 5 optometry specialists compared manual and automated cropping efficiency. Downstream task validation involved preprocessing 1000 primary gaze photographs using the Dlib toolkit, faster region-based convolutional neural network (R-CNN; both without head tilt correction), and our model (with correction), evaluating the impact of head tilt correction via the vision transformer strabismus screening network through 5-fold cross-validation. ResultsThe model achieved exceptional performance across datasets: on the 5-fold cross-validation set, it recorded a mean precision of 1.000 (95% CI 1.000‐1.000), recall of 1.000 (95% CI 1.000‐1.000), mAP50 of 0.995 (95% CI 0.995‐0.995), and mAP95 of 0.893 (95% CI 0.870‐0.918); on the internal independent test set, precision and recall were 1.000, with mAP50 of 0.995 and mAP95 of 0.801; and on the external cross-population test set, precision and recall were 1.000, with mAP50 of 0.937 and mAP95 of 0.792. The control experiment reduced image preparation time from 10 hours for manual cropping of 900 photos to 30 seconds with the automated model. Downstream strabismus screening task validation showed our model (with head tilt correction) achieving an area under the curve of 0.917 (95% CI 0.901‐0.933), surpassing Dlib-toolkit and faster R-CNN (both without head tilt correction) with an area under the curve of 0.856 (PP ConclusionsThis study delivers an AI-driven platform featuring a preprocessing algorithm that automates eye region cropping, correcting head tilt variations to improve image quality for AI development and clinical use. Integrated with electronic archives and patient-physician interaction, it enhances workflow efficiency, ensures telemedicine privacy, and supports ophthalmological research and strabismus care.
Audience Academic
Author Chen, Xiaohang
Wu, Dawen
Xie, Bin
Liu, Longqian
Li, Yanfei
Liu, Jingyu
Yang, Guoyuan
Shang, Wenyi
Yang, Zeyi
Zhang, Haixian
Yin, Teng
Author_xml – sequence: 1
  givenname: Dawen
  orcidid: 0000-0002-4938-3044
  surname: Wu
  fullname: Wu, Dawen
– sequence: 2
  givenname: Yanfei
  orcidid: 0009-0003-4563-7515
  surname: Li
  fullname: Li, Yanfei
– sequence: 3
  givenname: Zeyi
  orcidid: 0009-0007-2327-9082
  surname: Yang
  fullname: Yang, Zeyi
– sequence: 4
  givenname: Teng
  orcidid: 0009-0003-1254-6705
  surname: Yin
  fullname: Yin, Teng
– sequence: 5
  givenname: Xiaohang
  orcidid: 0000-0002-0165-3607
  surname: Chen
  fullname: Chen, Xiaohang
– sequence: 6
  givenname: Jingyu
  orcidid: 0009-0000-7153-397X
  surname: Liu
  fullname: Liu, Jingyu
– sequence: 7
  givenname: Wenyi
  orcidid: 0009-0001-5148-9358
  surname: Shang
  fullname: Shang, Wenyi
– sequence: 8
  givenname: Bin
  orcidid: 0009-0004-1144-9900
  surname: Xie
  fullname: Xie, Bin
– sequence: 9
  givenname: Guoyuan
  orcidid: 0000-0002-7314-836X
  surname: Yang
  fullname: Yang, Guoyuan
– sequence: 10
  givenname: Haixian
  orcidid: 0000-0002-9821-508X
  surname: Zhang
  fullname: Zhang, Haixian
– sequence: 11
  givenname: Longqian
  orcidid: 0000-0002-2265-4625
  surname: Liu
  fullname: Liu, Longqian
BackLink https://www.ncbi.nlm.nih.gov/pubmed/40674714$$D View this record in MEDLINE/PubMed
BookMark eNptkt1qFDEUx4NUbLv2FSQggjdbJx-TmfFG1m3VwkKL9eMyZJKT2dSZyTSZraxXvoMXvp9PYna3lhYkFwnn_PLjHPgfor3e94DQEcmOKanEq4LzjD5CB4SzclqWBdm7995HhzFeZRnNeEWeoH2eiYIXhB-g3ycAA16ACr3rmz8_f71VEQy-CKBddL7H8-CHIbWwt_h0DfgjNKkcsevx5RhU7WK3ivhi6UffBDUs42s8axsf3Ljs8AncQOuHDvoRq97gL6p1Ro0b7-W4MmtsfcBfffhmW_8dnw-j69yPbf8pemxVG-Ho9p6gz-9OP80_TBfn78_ms8VUc8HGqc1ZbYWuGDOEE6MzpYDz2hhWVoWlqoKsznNNOTBWsgoAiKKkpKbOjS2JZhN0tvMar67kEFynwlp65eS24EMjVRidbkEKk1W8rhWY3HJGVc2o5lyIImdcM2GT683ONazqDoxOWwfVPpA-7PRuKRt_IwmlFRNpwAl6eWsI_noFcZSdixraVvXgV1EyyggVghUsoc93aKPSbK63Pin1BpezkpeC5CzniTr-D5WOgc7plCDrUv3Bh2f3d7gb_l9iEvBiB-jgYwxg7xCSyU0S5TaJ7C9RatIQ
Cites_doi 10.1038/s41433-024-03228-5
10.1109/CVPR.2016.91
10.1007/978-3-319-46448-0_2
10.1109/ICCV.2017.322
10.1371/journal.pone.0255643
10.3760/cma.j.cn112142-20231111-00226
10.1016/s0161-6420(99)00035-4
10.2196/60226
10.3389/ebm.2024.10320
10.3390/make5040083
10.1109/ICCV.2015.169
10.1109/EHB47216.2019.8970033
10.1109/CVPR.2014.241
10.1186/s12911-021-01691-8
10.3390/diagnostics9030072
10.1109/ICCVW.2015.41
10.1136/bjo.2007.116905
10.1167/tvst.10.1.33
10.1007/s11263-019-01228-7
ContentType Journal Article
Copyright Dawen Wu, Yanfei Li, Zeyi Yang, Teng Yin, Xiaohang Chen, Jingyu Liu, Wenyi Shang, Bin Xie, Guoyuan Yang, Haixian Zhang, Longqian Liu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org).
COPYRIGHT 2025 Journal of Medical Internet Research
Copyright © Dawen Wu, Yanfei Li, Zeyi Yang, Teng Yin, Xiaohang Chen, Jingyu Liu, Wenyi Shang, Bin Xie, Guoyuan Yang, Haixian Zhang, Longqian Liu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org) 2025
Copyright_xml – notice: Dawen Wu, Yanfei Li, Zeyi Yang, Teng Yin, Xiaohang Chen, Jingyu Liu, Wenyi Shang, Bin Xie, Guoyuan Yang, Haixian Zhang, Longqian Liu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org).
– notice: COPYRIGHT 2025 Journal of Medical Internet Research
– notice: Copyright © Dawen Wu, Yanfei Li, Zeyi Yang, Teng Yin, Xiaohang Chen, Jingyu Liu, Wenyi Shang, Bin Xie, Guoyuan Yang, Haixian Zhang, Longqian Liu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org) 2025
DBID AAYXX
CITATION
CGR
CUY
CVF
ECM
EIF
NPM
7X8
5PM
DOA
DOI 10.2196/74402
DatabaseName CrossRef
Medline
MEDLINE
MEDLINE (Ovid)
MEDLINE
MEDLINE
PubMed
MEDLINE - Academic
PubMed Central (Full Participant titles)
DOAJ Directory of Open Access Journals
DatabaseTitle CrossRef
MEDLINE
Medline Complete
MEDLINE with Full Text
PubMed
MEDLINE (Ovid)
MEDLINE - Academic
DatabaseTitleList MEDLINE - Academic


MEDLINE

Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
– sequence: 2
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Medicine
Library & Information Science
EISSN 1438-8871
EndPage e74402
ExternalDocumentID oai_doaj_org_article_6d094bbaed5f432ab32c44667534c36f
PMC12293638
A848615354
40674714
10_2196_74402
Genre Validation Study
Journal Article
GeographicLocations Canada
GeographicLocations_xml – name: Canada
GroupedDBID ---
.4I
.DC
29L
2WC
36B
53G
5GY
5VS
77I
77K
7RV
7X7
8FI
8FJ
AAFWJ
AAKPC
AAWTL
AAYXX
ABDBF
ABIVO
ABUWG
ACGFO
ADBBV
AEGXH
AENEX
AFFHD
AFKRA
AFPKN
AIAGR
ALMA_UNASSIGNED_HOLDINGS
ALSLI
AOIJS
BAWUL
BCNDV
BENPR
CCPQU
CITATION
CNYFK
CS3
DIK
DU5
DWQXO
E3Z
EAP
EBD
EBS
EJD
ELW
EMB
EMOBN
ESX
F5P
FRP
FYUFA
GROUPED_DOAJ
GX1
HMCUK
HYE
IAO
ICO
IEA
IHR
INH
ISN
ITC
KQ8
M1O
M48
NAPCQ
OK1
OVT
P2P
PGMZT
PHGZM
PHGZT
PIMPY
PPXIY
PQQKQ
PRQQA
RNS
RPM
SJN
SV3
TR2
UKHRP
XSB
ACUHS
ALIPV
CGR
CUY
CVF
ECM
EIF
NPM
7X8
PUEGO
5PM
ID FETCH-LOGICAL-c463t-f53bf6c933d141dc0aae44bdd3897f2a9e0b55c24e33839eee1a2182db5df81c3
IEDL.DBID DOA
ISICitedReferencesCount 1
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001538507000001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1438-8871
1439-4456
IngestDate Mon Nov 10 04:35:47 EST 2025
Tue Nov 04 02:02:26 EST 2025
Fri Sep 05 15:39:02 EDT 2025
Tue Nov 11 10:45:01 EST 2025
Tue Nov 04 18:10:43 EST 2025
Wed Jul 30 01:32:59 EDT 2025
Sat Nov 29 07:42:52 EST 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 10
Keywords AI management system
image preprocessing
ocular alignment
artificial intelligence
clinical workflow
Language English
License Dawen Wu, Yanfei Li, Zeyi Yang, Teng Yin, Xiaohang Chen, Jingyu Liu, Wenyi Shang, Bin Xie, Guoyuan Yang, Haixian Zhang, Longqian Liu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org).
This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research (ISSN 1438-8871), is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c463t-f53bf6c933d141dc0aae44bdd3897f2a9e0b55c24e33839eee1a2182db5df81c3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ObjectType-Undefined-3
these authors contributed equally
ORCID 0000-0002-9821-508X
0000-0002-0165-3607
0009-0001-5148-9358
0009-0004-1144-9900
0009-0007-2327-9082
0000-0002-4938-3044
0009-0003-4563-7515
0009-0000-7153-397X
0009-0003-1254-6705
0000-0002-7314-836X
0000-0002-2265-4625
OpenAccessLink https://doaj.org/article/6d094bbaed5f432ab32c44667534c36f
PMID 40674714
PQID 3231266373
PQPubID 23479
ParticipantIDs doaj_primary_oai_doaj_org_article_6d094bbaed5f432ab32c44667534c36f
pubmedcentral_primary_oai_pubmedcentral_nih_gov_12293638
proquest_miscellaneous_3231266373
gale_infotracmisc_A848615354
gale_infotracacademiconefile_A848615354
pubmed_primary_40674714
crossref_primary_10_2196_74402
PublicationCentury 2000
PublicationDate 2025-07-17
PublicationDateYYYYMMDD 2025-07-17
PublicationDate_xml – month: 07
  year: 2025
  text: 2025-07-17
  day: 17
PublicationDecade 2020
PublicationPlace Canada
PublicationPlace_xml – name: Canada
– name: Toronto, Canada
PublicationTitle Journal of medical Internet research
PublicationTitleAlternate J Med Internet Res
PublicationYear 2025
Publisher Journal of Medical Internet Research
JMIR Publications
Publisher_xml – name: Journal of Medical Internet Research
– name: JMIR Publications
References Tan (R24); 21
Huang (R15); 16
Ünver (R23); 9
Wu (R6); 249
Zheng (R18); 10
King (R8)
Terven (R13); 5
Chia (R1); 91
R21
R20
R22
R25
Wu (R5); 38
Liu (R3); 60
Ming (R4); 26
R7
R9
Selvaraju (R14); 128
R10
Coats (R2); 107
R12
R11
R16
R17
R19
References_xml – volume: 38
  start-page: 3101
  issue: 16
  ident: R5
  article-title: An artificial intelligence platform for the screening and managing of strabismus
  publication-title: Eye (Lond)
  doi: 10.1038/s41433-024-03228-5
– ident: R19
  doi: 10.1109/CVPR.2016.91
– ident: R20
  doi: 10.1007/978-3-319-46448-0_2
– ident: R21
– ident: R17
  doi: 10.1109/ICCV.2017.322
– volume: 16
  issue: 8
  ident: R15
  article-title: An automatic screening method for strabismus detection based on image processing
  publication-title: PLoS One
  doi: 10.1371/journal.pone.0255643
– volume: 60
  start-page: 484
  issue: 6
  ident: R3
  article-title: Challenges and prospects in the application of artificial intelligence for ocular disease screening and diagnosis
  publication-title: Zhonghua Yan Ke Za Zhi
  doi: 10.3760/cma.j.cn112142-20231111-00226
– ident: R25
– volume: 107
  start-page: 402
  issue: 2
  ident: R2
  article-title: Impact of large angle horizontal strabismus on ability to obtain employment
  publication-title: Ophthalmology
  doi: 10.1016/s0161-6420(99)00035-4
– ident: R8
  publication-title: J Mach Learn Res
– volume: 26
  ident: R4
  article-title: Performance of ChatGPT in ophthalmic registration and clinical diagnosis: cross-sectional study
  publication-title: J Med Internet Res
  doi: 10.2196/60226
– volume: 249
  ident: R6
  article-title: Integrating artificial intelligence in strabismus management: current research landscape and future directions
  publication-title: Exp Biol Med (Maywood)
  doi: 10.3389/ebm.2024.10320
– ident: R10
– ident: R16
– volume: 5
  start-page: 1680
  issue: 4
  ident: R13
  article-title: A comprehensive review of YOLO architectures in computer vision: from YOLOv1 to YOLOv8 and YOLO-NAS
  publication-title: MAKE
  doi: 10.3390/make5040083
– ident: R9
  doi: 10.1109/ICCV.2015.169
– ident: R12
– ident: R22
  doi: 10.1109/EHB47216.2019.8970033
– ident: R7
  doi: 10.1109/CVPR.2014.241
– volume: 21
  issue: 1
  ident: R24
  article-title: Comparison of RetinaNet, SSD, and YOLO v3 for real-time pill identification
  publication-title: BMC Med Inform Decis Mak
  doi: 10.1186/s12911-021-01691-8
– volume: 9
  issue: 3
  ident: R23
  article-title: Skin lesion segmentation in dermoscopic images with combination of YOLO and GrabCut algorithm
  publication-title: Diagnostics (Basel)
  doi: 10.3390/diagnostics9030072
– ident: R11
  doi: 10.1109/ICCVW.2015.41
– volume: 91
  start-page: 1337
  issue: 10
  ident: R1
  article-title: Comitant horizontal strabismus: an Asian perspective
  publication-title: Br J Ophthalmol
  doi: 10.1136/bjo.2007.116905
– volume: 10
  start-page: 33
  issue: 1
  ident: R18
  article-title: Detection of referable horizontal strabismus in children’s primary gaze photographs using deep learning
  publication-title: Transl Vis Sci Technol
  doi: 10.1167/tvst.10.1.33
– volume: 128
  start-page: 336
  issue: 2
  ident: R14
  article-title: Grad-CAM: visual explanations from deep networks via gradient-based localization
  publication-title: Int J Comput Vis
  doi: 10.1007/s11263-019-01228-7
SSID ssj0020491
Score 2.451676
Snippet Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting artificial...
Background Traditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent, limiting...
Abstract BackgroundTraditional ocular gaze photograph preprocessing, relying on manual cropping and head tilt correction, is time-consuming and inconsistent,...
SourceID doaj
pubmedcentral
proquest
gale
pubmed
crossref
SourceType Open Website
Open Access Repository
Aggregation Database
Index Database
StartPage e74402
SubjectTerms Adult
Algorithms
Artificial Intelligence
Assistive Technology for Vision Loss/Impairment
Care and treatment
Cross-Sectional Studies
Deep Learning
Diagnosis
Eye - diagnostic imaging
Female
Health aspects
Humans
Image Processing, Computer-Assisted - methods
Machine Learning
Male
Medical screening
Methods
Middle Aged
Ophthalmology
Original Paper
Photography
Prospective Studies
Retrospective Studies
Strabismus
Strabismus - diagnostic imaging
Tools, Programs and Algorithms
Workflow
Title Deep Learning–Based Precision Cropping of Eye Regions in Strabismus Photographs: Algorithm Development and Validation Study for Workflow Optimization
URI https://www.ncbi.nlm.nih.gov/pubmed/40674714
https://www.proquest.com/docview/3231266373
https://pubmed.ncbi.nlm.nih.gov/PMC12293638
https://doaj.org/article/6d094bbaed5f432ab32c44667534c36f
Volume 27
WOSCitedRecordID wos001538507000001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: DOA
  dateStart: 19990101
  isFulltext: true
  titleUrlDefault: https://www.doaj.org/
  providerName: Directory of Open Access Journals
– providerCode: PRVPQU
  databaseName: Library Science Database
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: M1O
  dateStart: 20010101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/libraryscience
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Nursing & Allied Health Database
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: 7RV
  dateStart: 20010101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/nahs
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: ProQuest Central
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: BENPR
  dateStart: 20010101
  isFulltext: true
  titleUrlDefault: https://www.proquest.com/central
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: ProQuest Health & Medical Collection
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: 7X7
  dateStart: 20010101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/healthcomplete
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Publicly Available Content Database
  customDbUrl:
  eissn: 1438-8871
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0020491
  issn: 1438-8871
  databaseCode: PIMPY
  dateStart: 20010101
  isFulltext: true
  titleUrlDefault: http://search.proquest.com/publiccontent
  providerName: ProQuest
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV3NjtMwEB7BglZICMHyV9itjITgFG0SO03CrV26gkO7UQWrcooc22kjbZNV04K48Q4ceD-ehJnEXTXiwIWLD3YSTTxj-xt7_A3Aa-VmER8Y6eiMjhlFHDoSLclRRhjPJX511SabCKfTaD6Pk71UXxQT1tIDtx13OtDogGSZNDrIBfdlxn1FZ5AIs4Xig5xmXzeMd86UdbUQ93qHcJ8CndHETokFz--sPA1B_9_T8N461I2R3Ft0zh_CA4sW2bCV8hHcMuURnNi7BuwNs5eJqHOZHaVHcDix5-WP4dd7Y66Z5VBd_P7xc4SLlmbJ2mbWYWfriggaFqzK2fi7YTND8ck1K_B7KFBW1KttzZJltWmpret3bHi1qNbFZrliewFHTJaaXSKm11YYYq1lKByj3fj8qvrGLnByWtlbn0_g8_n409kHx6ZicJQY8I2TBzzLByrmXHvC08qV0giRaY14J8x9GRs3CwLlC0Mub2yM8SRxw-ss0HnkKf4UDsqqNM-BBXkg0BIQ-MgAvxBExvUl4ojQjXQsfdOD_k5N6XXLuJGip0J6TBs99mBEyrtpJILspgLNJrVmk_7LbHrwllSf0jDG7lTS3kZAGYkQKx1GIiIsHIgeHHeexOGnOs2vdsaTUhPFrJWm2tYpR-iM8IeHvAfPWmO6kRlhFO0G4NtRx8w6P9VtKYtlw_7t-YjQcNZ88T-64SXc8ymhccMUegwHm_XWnMBd9XVT1Os-3A5nl1TOw6aM-nBnNJ4ms34z2rCceBdYl3ycJF_-ADlcMuw
linkProvider Directory of Open Access Journals
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Deep+Learning%E2%80%93Based+Precision+Cropping+of+Eye+Regions+in+Strabismus+Photographs%3A+Algorithm+Development+and+Validation+Study+for+Workflow+Optimization&rft.jtitle=Journal+of+medical+Internet+research&rft.au=Wu%2C+Dawen&rft.au=Li%2C+Yanfei&rft.au=Yang%2C+Zeyi&rft.au=Yin%2C+Teng&rft.date=2025-07-17&rft.pub=Journal+of+Medical+Internet+Research&rft.issn=1439-4456&rft.volume=27&rft.issue=10&rft_id=info:doi/10.2196%2F74402&rft.externalDocID=A848615354
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1438-8871&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1438-8871&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1438-8871&client=summon