Sight + Sound, Spring 2020
by Carrie Fogel
Much has been discussed regarding the potential for what is termed ‘Big Data’ to revolutionize how we understand patterns and trends relative to human behavior and interactions. Big Data, loosely defined as extremely large data sets that can be analyzed computationally to systematically draw out information and form new understandings and predictions for trends, is being used in nearly every industry to be more precise and more efficient in the way business is done. The use of Big Data in healthcare has already led to many improvements, allowing physicians to provide more accurate insights about a disease and comprehensive recommendations for treatment.
The availability of electronic health records offers an unprecedented opportunity to apply predictive analytics to improve and change the practice of medicine. By leveraging scalable computational power, machine learning offers the potential to predict outcomes from real-world data extracted from electronic medical records. Recent studies have shown that computers can learn rich, hierarchical representations of raw data with much less human effort than previously required. Several successful examples of large-scale risk assessment models include hospital readmission models, disease onset prediction, and prediction of healthcare utilization and cost.
With ongoing support from the Shear Family Foundation, the Department of Ophthalmology has invested time and resources into building an artificial intelligence model that would predict the way that a disease, like macular degeneration, would progress in each individual and to allow us to test therapeutic interventions on a virtual patient or a ‘Digital Twin.’ The Digital Twin platform is a shared idea between José-Alain Sahel, MD, Distinguished Professor and Chairman of the Department of Ophthalmology at the University of Pittsburgh, along with clinician-scientist Kunal Dansingani, MBBS, MA, FRCOphth, Associate Professor of Ophthalmology, and Ryad Benosman, PhD, a robotic vision expert, as well as Jay Chabblani and Shan Suthaharan, PhD, a computer science professor, and Kiran Kumar Vupparaboina, PhD, an electrical engineer, that would use big data to show patients what the outcomes would be of the many different treatment options, be they drug therapies, surgery, or no intervention, and what the costs associated with each option would be. This, Dr. Sahel believes, is the best way for patients to make an informed choice about their care that fits best with their lives and abilities.
The Digital Twin project was first begun in Paris as part of an European project. It is now developed between the two sites with a strong team now established in Pittsburgh, working with experts at Carnegie Mellon University and UPMC. On of the goals is to build and tested novel machine-learning algorithms which learn from data acquired over disparate time scales and models and can predict disease progression. Dr. Dansingani now intends to move this model into clinical testing with data from real patients, with the hope that the model can function in real-time as an early detection and decision support tool for physicians, by modeling the potential clinical course of a range of diseases under various treatment conditions and by learning continuously from exposure to new case data. The team is working toward the development of a framework – the ‘Digital Twin’ platform – for an automated personalized medicine system, allowing for more effective treatment based on individual health data paired with predictive analytics.
Most recently, the team working on artificial intelligence in the Ophthalmology Department was awarded a generous grant from the Shear Family Foundation to continue improving the model, inputting data that has been collected from patients with macular degeneration and comparing the predictions given by the model to the actual disease pathology of the patient.
In the long term, as the algorithm is perfected and the model is tested in many different diseases beyond blindness, such as cancer, inflammatory diseases and others, we hope that the technology will put Pittsburgh on the map for being at the forefront of the Big Data health care revolution.