This content is only available for ASCRS Members
This content from the 2020 ASCRS Virtual Annual Meeting is only available to ASCRS members. To log in, click the teal "Login" button in the upper right-hand corner of this page.
Papers in this Session
Expand each tab below to view the paper abstract for each paper within this session.
A Novel Multi-Disease Artificial Intelligence Suite for the Autonomous Diagnosis of Dry Eye, Keratoconus and Fuchs Endothelial Dystrophy
Authors
Mohamed Abou Shousha, MD
Amr Elsawy, MSc
Taher K. Eleiwa, MD, MSc
Collin B. Chase
Eyup Ozcan, MD
Mohamed Y. Tolba, MD, PhD
Sonia H. Yoo, MD, ABO
Methods
A dataset of 46,179 ASOCT images (11,830 dry eye, 11,440 FED, 11,469 keratoconus and 11,440 normal controls images) was used to train and validate a deep learning AI model. The dataset was created from images of 736 patients obtained using ASOCT (Envisu R2210, Bioptigen, Buffalo Grove, IL), of which, 351 eyes were selected for having the highest quality scans and to create a balanced dataset. Diagnosis used in labeling was made by Bascom Palmer Eye Institute cornea specialists. The images were randomly divided into 80% training, 10% validation and 10% testing non-overlapped datasets. The classification accuracies and the area under receiver operating characteristic curve (AUC) were computed.
Results
Our AI system was able to achieve a training classification accuracy of 99.99%, a validation classification accuracy of 99.56%, and a testing classification accuracy of 99.51% in the diagnosis of dry eye, keratoconus and FED. We obtained accuracies of 99.2% for dry eye, 99.9% for keratoconus, 100% for FED, 98.7% for normal controls and overall accuracy of 99.51%. The AUC values were > 0.99 for all categories.
Conclusion
Our multi-disease AI diagnostic suite trained on ASOCT images is an accurate autonomous technique for the diagnosis of dry eye, keratoconus and Fuchs endothelial dystrophy. Validations in the setting of prospective multi-center clinical trials are warranted.
Mohamed Abou Shousha, MD
Amr Elsawy, MSc
Taher K. Eleiwa, MD, MSc
Collin B. Chase
Eyup Ozcan, MD
Mohamed Y. Tolba, MD, PhD
Sonia H. Yoo, MD, ABO
Methods
A dataset of 46,179 ASOCT images (11,830 dry eye, 11,440 FED, 11,469 keratoconus and 11,440 normal controls images) was used to train and validate a deep learning AI model. The dataset was created from images of 736 patients obtained using ASOCT (Envisu R2210, Bioptigen, Buffalo Grove, IL), of which, 351 eyes were selected for having the highest quality scans and to create a balanced dataset. Diagnosis used in labeling was made by Bascom Palmer Eye Institute cornea specialists. The images were randomly divided into 80% training, 10% validation and 10% testing non-overlapped datasets. The classification accuracies and the area under receiver operating characteristic curve (AUC) were computed.
Results
Our AI system was able to achieve a training classification accuracy of 99.99%, a validation classification accuracy of 99.56%, and a testing classification accuracy of 99.51% in the diagnosis of dry eye, keratoconus and FED. We obtained accuracies of 99.2% for dry eye, 99.9% for keratoconus, 100% for FED, 98.7% for normal controls and overall accuracy of 99.51%. The AUC values were > 0.99 for all categories.
Conclusion
Our multi-disease AI diagnostic suite trained on ASOCT images is an accurate autonomous technique for the diagnosis of dry eye, keratoconus and Fuchs endothelial dystrophy. Validations in the setting of prospective multi-center clinical trials are warranted.
Dry Eye Disease: A Canadian Quality of Life and Productivity Loss Survey Study
Authors
Clara C. Chan, MD, FRCS
Setareh Ziai, MD
Varun Myageri
James G. Burns, PhD
C. Lisa Prokopich, OD, MSc
Purpose
To describe the direct out-of-pocket costs, the indirect absenteeism and presenteeism costs, and the quality of life (QoL) impact attributable to dry eye disease (DED).
Methods
A multi-centre observational cross-sectional survey study. Patients were recruited during a routine consultation visit with their optometrist/ophthalmologist occurring between August 2018 and March 2019 across Canada. All consecutive patients who had a current diagnosis of DED by an accredited health care provider and met selection criteria requirements were invited to participate in the survey.
Results
Overall 151 patients fielded surveys, of which 146 patients were found to be eligible for the study. The results showed that dry eye disease was associated with annual direct patient costs of $2,324.35 and annual indirect patient costs (CAD) of $21,052.11. 101(69%) patients reported an Eye Dryness Score value >= 60, suggesting a high prevalence of severe cases. 104 patients (71%) were employed or seeking employment. The impact of DED on QoL and on both indirect and direct cost were associated with increased severity of DED and with the diagnosis of Sjögren Syndrome.
Conclusion
This study furthers the understanding on negative impact of DED on patients’ QoL and economic burden (including direct and indirect costs) based on severity of DED. Patients with severe DED reported lower QoL, higher medical fees, and impaired work productivity compared with mild and moderate cases.
Clara C. Chan, MD, FRCS
Setareh Ziai, MD
Varun Myageri
James G. Burns, PhD
C. Lisa Prokopich, OD, MSc
Purpose
To describe the direct out-of-pocket costs, the indirect absenteeism and presenteeism costs, and the quality of life (QoL) impact attributable to dry eye disease (DED).
Methods
A multi-centre observational cross-sectional survey study. Patients were recruited during a routine consultation visit with their optometrist/ophthalmologist occurring between August 2018 and March 2019 across Canada. All consecutive patients who had a current diagnosis of DED by an accredited health care provider and met selection criteria requirements were invited to participate in the survey.
Results
Overall 151 patients fielded surveys, of which 146 patients were found to be eligible for the study. The results showed that dry eye disease was associated with annual direct patient costs of $2,324.35 and annual indirect patient costs (CAD) of $21,052.11. 101(69%) patients reported an Eye Dryness Score value >= 60, suggesting a high prevalence of severe cases. 104 patients (71%) were employed or seeking employment. The impact of DED on QoL and on both indirect and direct cost were associated with increased severity of DED and with the diagnosis of Sjögren Syndrome.
Conclusion
This study furthers the understanding on negative impact of DED on patients’ QoL and economic burden (including direct and indirect costs) based on severity of DED. Patients with severe DED reported lower QoL, higher medical fees, and impaired work productivity compared with mild and moderate cases.
Contact Lens Association with Meibomian Gland Dysfunction in the Era of Meibography
Authors
Steven L. Carter, MD
Urmi V. Mehta
Priscilla Vu, MD, MD
Ryan G. Smith, MD
Alexander Knezevic, MD
Kailey A. Marshall, OD
Kavita K. Rao, MD, ABO
Marjan Farid, MD, ABO
Methods
A survey was given as part of standard workup for ocular surface disease at a tertiary academic center. The survey included type of contact lenses (CTL) used, frequency of use, duration of use, and if they slept in CTL. The survey also included the Standard Patient Evaluation of Eye Dryness (SPEED) questionnaire. Contact lens users were compared to non-contact lens users, and meibomian gland dropout and structural changes as evidenced on meibography was assessed.
Results
Initial 10 patients have not showed any significant results, suspected due to small sample size. Data will continue to be analyzed as it is obtained in the coming few weeks.
Conclusion
Our current hypothesized risk factors for meibomian gland dysfunction have not shown to have a significant impact, but sample size is small. Additional results and conclusions forthcoming.
Steven L. Carter, MD
Urmi V. Mehta
Priscilla Vu, MD, MD
Ryan G. Smith, MD
Alexander Knezevic, MD
Kailey A. Marshall, OD
Kavita K. Rao, MD, ABO
Marjan Farid, MD, ABO
Methods
A survey was given as part of standard workup for ocular surface disease at a tertiary academic center. The survey included type of contact lenses (CTL) used, frequency of use, duration of use, and if they slept in CTL. The survey also included the Standard Patient Evaluation of Eye Dryness (SPEED) questionnaire. Contact lens users were compared to non-contact lens users, and meibomian gland dropout and structural changes as evidenced on meibography was assessed.
Results
Initial 10 patients have not showed any significant results, suspected due to small sample size. Data will continue to be analyzed as it is obtained in the coming few weeks.
Conclusion
Our current hypothesized risk factors for meibomian gland dysfunction have not shown to have a significant impact, but sample size is small. Additional results and conclusions forthcoming.
Meibomian Gland Secretions in Patients with Meibographically Undetectable Glandular Contents
Authors
Jack V. Greiner, MD
Paula J. Oliver
Purpose
We have reported on the meibomian gland secretion (MGS) scores in patients with varying degrees of meibomian gland dysfunction (MGD) both diagnostically and therapeutically. This study presents data on MGS with manual expression of glands in patients with severe MGD and patients with meibographically undetectable/low glandular contents.
Methods
This observational study examined equal groups of patients (n=30) with MGD. All patients had SPEED scores ≥6 and meibomian gland assessment scores ≤12. Group 1 had meibography showing ̴50% gland area having secretions of a typical lipid color, whereas Group 2 had meibographic evidence of 0 to <10% gland area having such secretion. In Group 2 the majority of the glands appeared empty. With minimal conjunctival edema, glandular morphology could be discerned, though glands were transparent and without color. Meibomian gland evaluation was performed using the meibomian gland evaluator (MGE) (1.2 gm/mm2) followed by manual lid expression ( ̴1 kg/cm2). Secretions were graded on a 0-3 scale.
Results
The average±SE MGS for glands expressed using the MGE in Group 1 was 9.2±1.4 and in Group 2 2.8±0.8 (P <0.01) and by manual meibomian gland expression in Group 1 was 13.1±2.5 and in Group 2 6.6±2.1 (P <0.01). Although there was sparse evidence for gradable secretions from meibomian gland ostia when secretions were graded with the MGE (pressure equivalent to a forced blink) in both Groups due to the severity of MGD, when manual meibomian gland expression (many times more forceful) was performed there were clearly gradable secretions from both Groups. Secretions were observed in 13 of Group 2 patients even in some cases where there were 0 lipid containing glands observed meibographically.
Conclusion
The observation of secretion from meibomian gland ostia during meibomian gland expression when lipids are not detectable by meibography was unexpected. These observation and the presence of glandular outlines seen by meibography (without apparent lipid colored glandular fill) suggests that the meibomian glands are continuing to function.
Jack V. Greiner, MD
Paula J. Oliver
Purpose
We have reported on the meibomian gland secretion (MGS) scores in patients with varying degrees of meibomian gland dysfunction (MGD) both diagnostically and therapeutically. This study presents data on MGS with manual expression of glands in patients with severe MGD and patients with meibographically undetectable/low glandular contents.
Methods
This observational study examined equal groups of patients (n=30) with MGD. All patients had SPEED scores ≥6 and meibomian gland assessment scores ≤12. Group 1 had meibography showing ̴50% gland area having secretions of a typical lipid color, whereas Group 2 had meibographic evidence of 0 to <10% gland area having such secretion. In Group 2 the majority of the glands appeared empty. With minimal conjunctival edema, glandular morphology could be discerned, though glands were transparent and without color. Meibomian gland evaluation was performed using the meibomian gland evaluator (MGE) (1.2 gm/mm2) followed by manual lid expression ( ̴1 kg/cm2). Secretions were graded on a 0-3 scale.
Results
The average±SE MGS for glands expressed using the MGE in Group 1 was 9.2±1.4 and in Group 2 2.8±0.8 (P <0.01) and by manual meibomian gland expression in Group 1 was 13.1±2.5 and in Group 2 6.6±2.1 (P <0.01). Although there was sparse evidence for gradable secretions from meibomian gland ostia when secretions were graded with the MGE (pressure equivalent to a forced blink) in both Groups due to the severity of MGD, when manual meibomian gland expression (many times more forceful) was performed there were clearly gradable secretions from both Groups. Secretions were observed in 13 of Group 2 patients even in some cases where there were 0 lipid containing glands observed meibographically.
Conclusion
The observation of secretion from meibomian gland ostia during meibomian gland expression when lipids are not detectable by meibography was unexpected. These observation and the presence of glandular outlines seen by meibography (without apparent lipid colored glandular fill) suggests that the meibomian glands are continuing to function.
Screen Time Association with Meibomian Gland Dysfunction in the Era of Meibography
Authors
Ryan G. Smith, MD
Alexander Knezevic, MD
Priscilla Vu, MD, MD
Steven L. Carter, MD
Kailey A. Marshall, OD
Kavita K. Rao, MD, ABO
Marjan Farid, MD, ABO
Purpose
To understand the association between screen time with meibomian gland dysfunction and associated changes on meibography.
Methods
A survey was given as part of standard workup for ocular surface disease at a tertiary academic center. The survey included total hours of screen time (cell phones, tablets, and computers) as well as hours per day spent on average on each of the three screen types. Participants were also asked if their occupation required screen usage. The survey also included the Standard Patient Evaluation of Eye Dryness (SPEED) questionnaire. Relationships between survey results and meibomian gland dropout and structural changes as evidenced on meibography were assessed.
Results
TBD
Conclusion
TBD
Ryan G. Smith, MD
Alexander Knezevic, MD
Priscilla Vu, MD, MD
Steven L. Carter, MD
Kailey A. Marshall, OD
Kavita K. Rao, MD, ABO
Marjan Farid, MD, ABO
Purpose
To understand the association between screen time with meibomian gland dysfunction and associated changes on meibography.
Methods
A survey was given as part of standard workup for ocular surface disease at a tertiary academic center. The survey included total hours of screen time (cell phones, tablets, and computers) as well as hours per day spent on average on each of the three screen types. Participants were also asked if their occupation required screen usage. The survey also included the Standard Patient Evaluation of Eye Dryness (SPEED) questionnaire. Relationships between survey results and meibomian gland dropout and structural changes as evidenced on meibography were assessed.
Results
TBD
Conclusion
TBD