An Automated and Efficient Glaucoma Prescreening System
Primary Information
Domain
Healthcare
Project No.
6013
Sanction and Project Initiation
Sanction No: F.No.-35-10/2016-TS-I dtd 27.2.17 (P ID 6013)
Sanction Date: 00/00/0000
Project Initiation date: 00/00/0000
Project Duration: 24
Partner Ministry/Agency/Industry
Indian Council of Medical Research
Role of partner:Financial grant
Support from partner:
Principal Investigator
Chandra Sekhar Seelamantula
Indian Institute of Science, Bangalore
Host Institute
Co-PIs
Dr. Prasanta Kumar Ghosh
Indian Institute of Science, Bangalore
Dr. Yogish S. Kamath
Department of Ophthalmology, Kasturba Medical College Hospital, Manipal, 576104
Scope and Objectives
1. Development of robust image processing algorithms to perform optic disc and optic cup outlines. 2. Extract relevant parameters (in addition to cup-to-disc ratio) that are specific to glaucoma. 3. Development of classification rules or algorithms to assess various glaucomatous conditions. 4. Development of the hardware for fundus image acquisition in the mydriatic mode. 5. Development of the image processing software for Android and iOS smartphone/handheld devices. 6. Testing and validation with the help of doctors and industry partners.
Deliverables
1. Software consisting of algorithms for automatic glaucoma prescreening using commercially available fundus imaging devices with desktop/laptop and smartphones/notebooks. 2. A detailed document explaining the methodology for deploying, maintaining and testing the whole system. 3. Publications in premier journals and top-tier conferences. 4. Technology transfer or liaison with industry 5. Creation of a fundus image database (at least 500 images) with expert annotation
Videos
Not yet
Scientific Output
Research Output 10 papers have been presented at top tier peer-reviewed international conferences and 2 more submissions are underway in premier journals. A new disease specific parameter developed for a more optimal classification of the glaucomatous condition. Software The performance of the proposed technique has been validated on several retinal fundus image databases varying in size, resolution, contrast, and illumination. The publicly available databases used are Messidor, Drishti-GS, DiaretDB, Drive, and Drions-DB, containing 1200, 101, 89, 40, and 110 retinal fundus images, respectively. The locally obtained databases are from Forus Health Pvt. Ltd. and Bosch Eye Care Solutions containing 126 and 60 retinal fundus images, respectively. This amounts to a total of 1726 images upon which the solution was validated. Field Trial and data collection: 1.Ethical committee clearance was obtained from the Institutional Ethics Committee of Kasturba Medical College (KMC), Manipal, to conduct field trials. 2.Remidio Innovative Solutions Pvt. Ltd. has been given 10 licenses for a six-month period for conducting field trials independently and passing on feedback. Competitions Finalist for the Bangalore Chamber of Commerce Innovation Day competition Finalist at GYTI BIRAC-SRISTI Award Won the Digital Health Prize at the Department of Biotechnology National Bio-Entrepreneurship Competition Grand Challenge 2018.
Results and outcome till date
A fully automated glaucoma-prescreening tool based on clinically relevant parameters has been implemented. ICD-9 rules based on CDR and DDLS based on RDR has been employed for the severity/stage analysis of glaucoma. ImageJ plugins and macros based on Java have been developed for individual/mass handling of fundus image data, respectively. The algorithm has been tested on various publicly available and privately obtained fundus image databases. The algorithm is proved to be efficient and accurate. We showed that the accuracy of prescreening is 8% to 10% higher when one uses the RDR-based DDLS instead of the frequently employed CDR. From a clinical perspective also, the RDR is more representative of the severity of the glaucomatous stage than the CDR as it captures the loss of neuroretinal fiber more accurately. The high accuracy is reflected with both two-stage and three-stage glaucoma classification. As far as estimating the parameter is concerned, we developed a novel active disc template based methodology operating in a multiresolution framework for reasons of computational efficiency. The optic disc and cup segmentation results were found to be in high agreement with the expert outlines over a large number of images from various databases. In keeping with the contemporary and game-changing trend of developing smartphone-based eyecare solutions, we have also developed associated software in Java, and Apps for Android and iOS devices, which operate in real-time on the device without the need for processing on a centralized server or in the cloud. Under the non-disclosure agreement signed with Remidio Innovative Solutions Pvt. Ltd., Bangalore, an iOS based app has been handed over for testing. The same was showcased in the recently concluded All India Ophthalmological Society (AIOS) meeting held at Coimbatore, India and at the Delhi Ophthalmology Conference 2018. Talks are also underway with other industry competitors such as IntuVision Labs Pvt. Ltd., Rayd8, and ChironX for incorporation of our apps and algorithms into their fundus image acquisition system. In parallel, with clearances from the Ethics Committee of Kasturba Medical College, Manipal Academy of Higher Education, Manipal to acquire patient data to carry out further validations, clinical validation is underway at the Department of Ophthalmology, Kasturba Medical College since March 2018. Thus far, we have been able to acquire and validate the performance of the solution on about eight hundred fundus images. About 1500 images have been collected so far. We have also refined the user interface in Android and iOS implementations based on feedback received from various sources. We have also optimized the functionality of the App for faster analysis. We are also rigorous testing and optimization of the workflow in the App. Further, the iOS and Android implementations are in sync.
Societal benefit and impact anticipated
Glaucoma is the 4th largest cause of blindness in India, and the 2nd leading cause in the world. Over 60 million suffering worldwide of which only 50% are aware. A 58% increase is expected over the next decade. The condition is found to be increasingly common beyond the age of 35. As per a census report presented by the Glaucoma Society of India in 2013, it is estimated that about 90% of the cases are left undiagnosed until too late in developing and third world countries. As a result, glaucoma is fast gaining the status of a potential epidemic. Thus, definitive point-of-care screening infrastructure in high demand. These software solutions are particularly attractive for triaging applications using modern day handheld fundus imaging devices, which are heavily deployed for mass screening purposes in the third-world countries in particular those in south Asia and Africa. Glaucoma prescreening devices are fast becoming smartphone-centric and are in need of efficient solutions for performing image analysis. Our invention precisely fills the gap. Our solution is versatile -- it goes well with a desktop computer as well as an Android/iOS device. This would enable a great deal of deskilling and would result in a fast and reliable diagnosis, which is particularly needed in the rural setting where prescreening camps are conducted from time to time.
Next steps
The clinical data collection part is in progress at Kasturba Medical College, Manipal. Subsequently, we shall perform clinical validation on that dataset. About 1500 images have been collected so far. We are also planning to outsource the user interface part of the App development work to a professional App developer to improve the visual appeal.
Publications and reports
J. R. Harish Kumar, A. K. Pediredla, and C. S. Seelamantula, Active discs for automated optic disc segmentation in Proc. IEEE Global Conference on Signal and Information Processing, pp. 225-229, 2015. J. R. Harish Kumar, S. Harsha, Y. Kamath, R. Jampala, and C. S. Seelamantula, Automatic optic cup segmentation using Kasa circle fitting technique, in Proc. IEEE Region Ten Conference, pp. 25-30, 2017. J. R. Harish Kumar, S. Sachi, K. Chaudhury, S. Harsha, and B. K. Singh, A unified approach for detection of diagnostically significant regions-of-interest in retinal fundus images, in Proc. IEEE Region Ten Conference, pp. 19-24, 2017. Dhruv Mohan, J. R. Harish Kumar, Chandra Sekhar Seelamantula, High-performance optic disc segmentation using convolutional neural networks, presented at IEEE International Conference on Image Processing, 2018. Subramanya Jois S. P., Harsha S. and J. R. Harish Kumar, Automatic Optic Disc Localization Using Particle Swarm Optimization Technique, presented at IEEE Region Ten Conference, 2018. Harsha S., J. R. Harish Kumar, Subramanya Jois S.P., and Chandra Sekhar Seelamantula, An Unconstrained Ellipse Fitting Technique And Application To Optic Cup Segmentation, presented at IEEE GlobalSIP 2018.
Patents
J. R. Harish Kumar and C. S. Seelamantula, Automated fundus image processing techniques for glaucoma prescreening, applied for Indian (Ref. no: 201741017981; App. no: TEMP/E-1/18223/2017-CHE) and International patent (New PCT International Application No. PCT/IB2018/053030)
Scholars and Project Staff
Nagaraj Adiga Research Associate June 1 - 30, 2017 Aishwarya Selvaraj Project Assistant July 2017 - August 2017 Aritra Bhowmik Project Associate July 2017 - November 2017 Harsha Sridhar Project Assistant September 2017 - February 2018 Subramanya Jois Project Assistant February 2018 - February 2018 Aniketh .M Project Assistant July 2018 - February 2018 Ashish Rao .M Project Associate August 2018 - February 2018 Sukriti Paul Project Assistant October 2018 - February 2018
Challenges faced
None
Other information
Won the Digital Health award at NBEC 2018
Financial Information
-
Total sanction: Rs. 6145000.00
-
Amount received: Rs. 4818250.00
-
Amount utilised for Equipment: Rs. 1100000.00
-
Amount utilised for Manpower: Rs. 1430039.00
-
Amount utilised for Consumables: Rs. 226882.00
-
Amount utilised for Contingency: Rs. 224500.00
-
Amount utilised for Travel: Rs. 140184.00
-
Amount utilised for Other Expenses: 54355.00
-
Amount utilised for Overheads: Rs. 803150.00
Equipment and facilities
Remidio fundus-on-phone device Volk inview fundus-on-phone device iPhone and Android phones 5 kVA UPS (partial) GPU System
