From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning
Anterior chamber depth (ACD) is a major risk factor of angle closure disease, and has been used in angle closure screening in various populations. However, ACD is measured from ocular biometer or anterior segment optical coherence tomography (AS-OCT), which are costly and may not be readily availabl...
| Main Authors: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2023-02-01
|
| Series: | PLOS Digital Health |
| Online Access: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9931242/?tool=EBI |
| _version_ | 1827849331680477184 |
|---|---|
| author | Zhi Da Soh Yixing Jiang Sakthi Selvam S/O Ganesan Menghan Zhou Monisha Nongiur Shivani Majithia Yih Chung Tham Tyler Hyungtaek Rim Chaoxu Qian Victor Koh Tin Aung Tien Yin Wong Xinxing Xu Yong Liu Ching-Yu Cheng |
| author_facet | Zhi Da Soh Yixing Jiang Sakthi Selvam S/O Ganesan Menghan Zhou Monisha Nongiur Shivani Majithia Yih Chung Tham Tyler Hyungtaek Rim Chaoxu Qian Victor Koh Tin Aung Tien Yin Wong Xinxing Xu Yong Liu Ching-Yu Cheng |
| author_sort | Zhi Da Soh |
| collection | DOAJ |
| description | Anterior chamber depth (ACD) is a major risk factor of angle closure disease, and has been used in angle closure screening in various populations. However, ACD is measured from ocular biometer or anterior segment optical coherence tomography (AS-OCT), which are costly and may not be readily available in primary care and community settings. Thus, this proof-of-concept study aims to predict ACD from low-cost anterior segment photographs (ASPs) using deep-learning (DL). We included 2,311 pairs of ASPs and ACD measurements for algorithm development and validation, and 380 pairs for algorithm testing. We captured ASPs with a digital camera mounted on a slit-lamp biomicroscope. Anterior chamber depth was measured with ocular biometer (IOLMaster700 or Lenstar LS9000) in data used for algorithm development and validation, and with AS-OCT (Visante) in data used for testing. The DL algorithm was modified from the ResNet-50 architecture, and assessed using mean absolute error (MAE), coefficient-of-determination (R2), Bland-Altman plot and intraclass correlation coefficients (ICC). In validation, our algorithm predicted ACD with a MAE (standard deviation) of 0.18 (0.14) mm; R2 = 0.63. The MAE of predicted ACD was 0.18 (0.14) mm in eyes with open angles and 0.19 (0.14) mm in eyes with angle closure. The ICC between actual and predicted ACD measurements was 0.81 (95% CI 0.77, 0.84). In testing, our algorithm predicted ACD with a MAE of 0.23 (0.18) mm; R2 = 0.37. Saliency maps highlighted the pupil and its margin as the main structures used in ACD prediction. This study demonstrates the possibility of predicting ACD from ASPs via DL. This algorithm mimics an ocular biometer in making its prediction, and provides a foundation to predict other quantitative measurements that are relevant to angle closure screening. Author summary This proof-of-concept study aimed to predict anterior chamber depth (ACD) quantitatively from anterior segment photographs (ASPs) using deep-learning. Anterior chamber depth is a major and consistent risk factor of primary angle closure disease (PACD), which is a major cause of glaucoma-induced blindness at the later stages. Our study is motivated by the lack of an appropriate screening tool for PACD, where clinical tests lack repeatability and imaging devices are too costly to adopt in primary care and community settings for screening. In this study, we included 2,311 pairs of ASPs and ACD measurements in algorithm development, and 380 pairs for algorithm testing. We modified the ResNet-50 convolutional neural network (CNN) in developing our algorithm. Our algorithm was able to predict ACD with a mean absolute error of 0.18mm in algorithm validation, and 0.23mm in algorithm testing. We obtained an intraclass correlation coefficient (ICC) of 0.81 in validation, indicating good agreement between actual and predicted ACD measurements. Importantly, our algorithm highlighted the pupil and its margins in predicting ACD, which is similar to how actual measurements were obtained from imaging devices. Our study shows that depth (i.e., ACD) may be predicted from 2-dimension photographs using deep-learning. This algorithm provides a foundation for predicting other relevant parameters of PACD, which may be further combined into a screening algorithm for the disease at a later stage. |
| first_indexed | 2024-03-12T10:02:41Z |
| format | Article |
| id | doaj.art-b318d9085d4c4b3d9213612d3b318af3 |
| institution | Directory Open Access Journal |
| issn | 2767-3170 |
| language | English |
| last_indexed | 2024-03-12T10:02:41Z |
| publishDate | 2023-02-01 |
| publisher | Public Library of Science (PLoS) |
| record_format | Article |
| series | PLOS Digital Health |
| spelling | doaj.art-b318d9085d4c4b3d9213612d3b318af32023-09-02T11:33:32ZengPublic Library of Science (PLoS)PLOS Digital Health2767-31702023-02-0122From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learningZhi Da SohYixing JiangSakthi Selvam S/O GanesanMenghan ZhouMonisha NongiurShivani MajithiaYih Chung ThamTyler Hyungtaek RimChaoxu QianVictor KohTin AungTien Yin WongXinxing XuYong LiuChing-Yu ChengAnterior chamber depth (ACD) is a major risk factor of angle closure disease, and has been used in angle closure screening in various populations. However, ACD is measured from ocular biometer or anterior segment optical coherence tomography (AS-OCT), which are costly and may not be readily available in primary care and community settings. Thus, this proof-of-concept study aims to predict ACD from low-cost anterior segment photographs (ASPs) using deep-learning (DL). We included 2,311 pairs of ASPs and ACD measurements for algorithm development and validation, and 380 pairs for algorithm testing. We captured ASPs with a digital camera mounted on a slit-lamp biomicroscope. Anterior chamber depth was measured with ocular biometer (IOLMaster700 or Lenstar LS9000) in data used for algorithm development and validation, and with AS-OCT (Visante) in data used for testing. The DL algorithm was modified from the ResNet-50 architecture, and assessed using mean absolute error (MAE), coefficient-of-determination (R2), Bland-Altman plot and intraclass correlation coefficients (ICC). In validation, our algorithm predicted ACD with a MAE (standard deviation) of 0.18 (0.14) mm; R2 = 0.63. The MAE of predicted ACD was 0.18 (0.14) mm in eyes with open angles and 0.19 (0.14) mm in eyes with angle closure. The ICC between actual and predicted ACD measurements was 0.81 (95% CI 0.77, 0.84). In testing, our algorithm predicted ACD with a MAE of 0.23 (0.18) mm; R2 = 0.37. Saliency maps highlighted the pupil and its margin as the main structures used in ACD prediction. This study demonstrates the possibility of predicting ACD from ASPs via DL. This algorithm mimics an ocular biometer in making its prediction, and provides a foundation to predict other quantitative measurements that are relevant to angle closure screening. Author summary This proof-of-concept study aimed to predict anterior chamber depth (ACD) quantitatively from anterior segment photographs (ASPs) using deep-learning. Anterior chamber depth is a major and consistent risk factor of primary angle closure disease (PACD), which is a major cause of glaucoma-induced blindness at the later stages. Our study is motivated by the lack of an appropriate screening tool for PACD, where clinical tests lack repeatability and imaging devices are too costly to adopt in primary care and community settings for screening. In this study, we included 2,311 pairs of ASPs and ACD measurements in algorithm development, and 380 pairs for algorithm testing. We modified the ResNet-50 convolutional neural network (CNN) in developing our algorithm. Our algorithm was able to predict ACD with a mean absolute error of 0.18mm in algorithm validation, and 0.23mm in algorithm testing. We obtained an intraclass correlation coefficient (ICC) of 0.81 in validation, indicating good agreement between actual and predicted ACD measurements. Importantly, our algorithm highlighted the pupil and its margins in predicting ACD, which is similar to how actual measurements were obtained from imaging devices. Our study shows that depth (i.e., ACD) may be predicted from 2-dimension photographs using deep-learning. This algorithm provides a foundation for predicting other relevant parameters of PACD, which may be further combined into a screening algorithm for the disease at a later stage.https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9931242/?tool=EBI |
| spellingShingle | Zhi Da Soh Yixing Jiang Sakthi Selvam S/O Ganesan Menghan Zhou Monisha Nongiur Shivani Majithia Yih Chung Tham Tyler Hyungtaek Rim Chaoxu Qian Victor Koh Tin Aung Tien Yin Wong Xinxing Xu Yong Liu Ching-Yu Cheng From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning PLOS Digital Health |
| title | From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning |
| title_full | From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning |
| title_fullStr | From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning |
| title_full_unstemmed | From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning |
| title_short | From 2 dimensions to 3rd dimension: Quantitative prediction of anterior chamber depth from anterior segment photographs via deep-learning |
| title_sort | from 2 dimensions to 3rd dimension quantitative prediction of anterior chamber depth from anterior segment photographs via deep learning |
| url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9931242/?tool=EBI |
| work_keys_str_mv | AT zhidasoh from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT yixingjiang from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT sakthiselvamsoganesan from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT menghanzhou from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT monishanongiur from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT shivanimajithia from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT yihchungtham from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT tylerhyungtaekrim from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT chaoxuqian from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT victorkoh from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT tinaung from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT tienyinwong from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT xinxingxu from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT yongliu from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning AT chingyucheng from2dimensionsto3rddimensionquantitativepredictionofanteriorchamberdepthfromanteriorsegmentphotographsviadeeplearning |