US Patent 9968257 Volumetric quantification of cardiovascular structures from medical imaging

A computing system accesses complementary image data of a biological tissue structure (BTS), which can include a human cardiovascular structure. The complementary image data is comprised of two-dimensional images which represent different views of the BTS respectively aligned with a plurality of distinct image planes. A plurality of separate convolutional neural networks (CNNs) are used to respectively process each of the plurality of two-dimensional images. Each CNN determines a probability map which is then adaptively fused into a single segmented output. A contouring operation is automatically performed to calculate at least one clinical measurement and/or create at least one 3D volume.