In the traditional industry model, even the slightest design change has to go through the entire optimization process. Once the parametric geometry is created, an analysis is run, depending on the various inputs, boundary conditions and the desired output values. These outputs are studied and optimized against the input values using either traditional data analytics or by analysis of testing results. Based on the results of this analysis, the design parameters and inputs are changed as desired and the optimization cycle is repeated. Such industries need methods to reduce this time by a significant amount. The motivation of this study is to create an algorithm that learns the structural analysis of a bioprosthetic heart valve using a convolutional neural network, which on being trained can predict the results of the analysis given any new geometry without going through the process of simulation, which is time consuming and may not be feasible without appropriate resources at all times. This is achieved by exploiting the parametric non-uniform roational B-splines (NURBS) formulation of the heart valve geometry through isogeometric analysis. The initial and final deformations of about 18000 simulations are provided as training data to a deep convolutional neural network. Key parameters of interest such as coaptation area are predicted and all results are validated against the results obtained from traditional structural analysis.