Automated Image Analysis
The Medalytix iGrading platform is an environment which can call and run automated computer algorithms for the early detection of disease risk within digital images.
iGradingM
Systematic screening for diabetic retinopathy using digital retinal photography has been shown to reduce the incidence of blindness amongst people with diabetes. There are many challenges to the implementation of widespread screening and use of automation is one way to improve screening efficiency and ensure consistency and accuracy of grading review.
Current approaches to automated detection of retinopathy have now reached the limit of human intra-reader variability. iGradingM is the only commercially available automated diabetic retinopathy screening platform to have been validated within an organised, systematic screening programme. Deployment of iGradingM within nascent diabetic retinopathy screening programmes ensures healthcare providers can be confident that the initial review of their patient’s images is consistent with UK NHS gold standard grading practice.
Medalytix iGradingM software is a stand‐alone medical device supplied as an off the shelf product for integration into existing hardware systems in the clinical setting. The software consists of an image analysis algorithm, assessing image quality and signs of diabetic retinopathy, integrated in a web‐based application framework which includes a graphical user interface and relational database. The software enables a user to load and process one or more images sourced from:
- an external system where the images have been assembled as part of a patient episode and are received and processed by the software over a local or wide area network; or
- from an external memory device where the images have been loaded directly into a dedicated folder on the host platform file system.
iGradingM provides the user with the facility to monitor the progress and status of analysis of images loaded into the system and generates reports on the subsequent image analysis results. No images are retained by the software, however the results of all images processed are maintained in a database which enables the user to review previous image analysis reports and produce tailored reports based on specific date ranges.
Use of iGradingM allows:
- Validation of image quality for diabetic retinopathy screening;
- Accurate and objective detection of the presence of microaneurysms;
- Management of diabetic retinopathy screening workload through filtration of retinal images and identification of those which have signs of disease
- Significant workload reduction
iGradingM has been tried, tested and validated against an unparalleled number of data sets and the algorithms deployed within it are the subject of many publications in peer-reviewed journals by the academic and clinical development team at the University of Aberdeen and NHS Grampian. The multidisciplinary team in Aberdeen have been involved in the optimisation of diabetic retinopathy screening practice for over 14 years. Key publications supporting the current algorithms are cited below:
- Automated grading for diabetic retinopathy: a large‐scale audit using arbitration by clinical experts. Fleming AF, Goatman KA, Philips S, Prescott GL, Sharp PF & Olson JA. Br J Ophthalmol 2010; 94:1606‐1610
- The efficacy of automated “disease/no disease” grading for diabetic retinopathy in a systematic screening programme. Philip S, Fleming AD, Goatman KA, et al. Br J Ophthalmol 2007; 91: 1512‐1517
- Automated assessment of diabetic retinal image quality based on clarity and field definition. Fleming AD, Philip S, Goatman KA, Olson JA & Sharp PF. Invest Ophthalmol Vis Sci 2006;47:1120‐1125
- Automated microaneurysm detection using local contrast normalisation and local vessel detection. Fleming AD, Philip S, Goatman KA, Olson JA & Sharp PF. IEEE Trans Med Imaging2006;25(9):1223‐1232
iGradingCVD
Medalytix is currently expanding the application of the iGradingM core algorithms into the cardiovascular disease (CVD) risk prediction field. Using a novel, patented prediction algorithm in-licensed from the Centre for Eye Disease Australia (CERA), iGradingCVD is currently being validated in clinical studies in the UK, and has the potential to provide a simple, non-invasive alternative to prediction of cardiovascular disease risk via a simple digital retinal image.
iGradingCVD is unique in the field and combines two different algorithmic approaches from two world-leading groups in their respective areas of expertise – the University of Aberdeen (retinal image analysis) and CERA (CVD risk prediction).
The algorithms for automated detection of retinopathy incorporated within the iGradingCVD product are the most extensively validated in the field and are the only ones that have been clinically validated within a live screening programme.
The CVD risk prediction algorithm has been developed by the academic and clinical team at CERA in Melbourne. This team are a leading group in the epidemiology of retinal assessment and cardiovascular risk, and the algorithm has been developed using the largest database of epidemiological studies relating cardiovascular risk and retinal features in the world.
Risk prediction engines are used in clinical practice to predict a person’s risk of having an adverse cardiovascular event, such as a stroke or a heart attack. These prediction engines typically combine a number of parameters including age, gender, smoking, blood pressure, weight/ BMI, blood lipid levels, whether diabetes is present or not, whether there is a family history or not, but always require a clinical investigation and a blood test of some description. These involve significant time and money. iGradingCVD is a new, non-invasive approach to assessment of CVD risk using an automated retinal analysis which performs with equivalent predictive ability to current practice.
The combination of these two world-leading approaches makes iGradingCVD a unique and powerful tool in the non-invasive, automated detection of CVD risk.