Diabetic retinopathy remains one of the leading causes for vision impairment among working adults. The goal of this paper is to employ variational data assimilation to achieve early detection so as to prevent the condition from being irreversible. Method: The forward computational fluid dynamics simulation adopted by previous research has been proven to be insufficient due to the non-patient-specific boundary conditions. A PDE constrained optimisation approach is proposed to infer the blood flow information via manipulating inlets/outlets boundary conditions as control variables. Results: The assimilated solutions were compared to the in vivo results. It is shown a better agreement is achieved. Conclusion: The study demonstrates the variational data assimilation is a feasible methodology to assist diabetic retinopathy early detection. Significance: The approach enables the definition of the patient-specific models based on a small number of particle velocimetry experimental measurements which crucially need not be obtained at the vessels corresponding to the inlets/outlets.
Watch seminar via Zoom: https://unsw.zoom.us/j/98279879459
Tue, 05/05/2020 - 11:05am