3D medical objects processing and retrieval using Spherical Harmonics: a case study with Congestive Heart Failure MRI exams

Abstract

Spherical Harmonics (SPHARMs) have been widely used in the three-dimensional (3D) object processing domain. The harmonic coefficients generated by this mathematical theory are considered a robust source of information about 3D objects analyzed. This information is used for different purposes like 3D modeling, lighting, and objects description. Some works already use SPHARMs to compare 3D objects, but their application in the medical object retrieval domain is innovative. This work presents the use of SPHARMs to aid the diagnosis of Congestive Heart Failure (CHF) disease, by retrieving similar cases, given a 3D model of the heart as a query argument. After implementing SPHARMs using 3D objects reconstructed from Magnetic Resonance Imaging exams, we validated our approach by executing retrievals from objects with and without CHF. The results indicated an average precision of 80%. In addition, the execution time was 60% lower than some descriptors previously tested. Robustness of SPHARMs in a specific application domain is corroborated, showing that they can be a promising descriptor for 3D medical objects.