3D heart reconstruction

Abstract

The purpose of this thesis was to achieve a 3D reconstruction of the four heart chambers using 2D echocardiographic images. A level set algorithm based on the phase symmetry approach and on a new logarithmic based stopping function was used to extract simultaneously the four heart cavities from these images in a fully automatic way. However to proceed to the 3D reconstruction using the segmented images, it was first necessary to satisfy clinical practise requirements. This means that the algorithm had to be validated to access the performance of the segmentation. Regarding this, the framework of this thesis was divided in two parts: validation of the segmentation algorithm and 3D reconstruction. The contours obtained in the segmentation were compared with the ones obtained by four physicians to evaluate the performance, reliability and confidence for eventual clinical practice. That algorithm evaluation versus clinicians‟ performance was made using eleven figures of merit: Mean/Max/Larger than 5 pixels Distance, Pratt Function, Hausdorff Distance, Similarity Angle, Similarity Region, Accuracy, Overlap, Sensitivity and Specificity; and two statistical tools: Box-plots and Dendrograms. The results indicate a reliable performance of the level set method for all chambers. The evaluation was based on echocardiography images of children. The 3D reconstruction was achieved using a heart phantom. This phantom was mainly composed by four balloons attached together and submerged in a water environment. A robotic arm with an ultrasound probe attached was used to take a large number of frames from the heart phantom. Several attempts were made using two types of acquisition: in Rotation and in Translation. Offline reconstructions of two rotations and one translation were presented and analyzed. The results of the rotations were far better than the translation. It was possible to infer the shape and volume of the balloons. These results present one more step in the way for a real-time 3D reconstruction using a tele-echographic system