A.4-Medical Imaging


CT scan, MRI, PET scan, 3D ultrasound… all these names are common place in hospitals and many visits to the doctor, and they all imply thousands of voxels to represent structure and/or function of the human body in 3D images. The amount of data acquired at medical facilities is growing exponentially and it is important to improve computational tools in order to help doctors diagnose patients from this data more efficiently. From the computational standpoint, the main challenge is the performance reliability required in such scenario: the algorithm has to guarantee an extremely low probability of false negatives, since patient diagnostics might depend on it. Moreover, the variability in the data from different patients, different machines and different technicians is very large, so large numbers of datasets are needed to test new algorithms.

Figure 1:  Example of 3D ultrasound images to analyze fetal anatomy. (Upper left) Schematic of axial views of the fetal head: (a) Transventricular plane; (b) transthalamic plane; (c) transcerebellar plane. Each figure indicates anatomies of interest to follow the correct development of a human embryo (From [1]).




[1] G. Carneiro, F. Amat, B. Georgescu, S. Good, and D. Comaniciu, “Semantic-based indexing of fetal anatomies from 3-D ultrasound data using global/semi-local context and sequential sampling,” in 2008 CVPR, Anchorage, USA, June 2008, pp. 1-8.