Quantitative analysis framework for SPECT·CT Tc-99m bone scintigraphy

Clinical SPECT routine uses tedious manual image evaluation using simplistic region of interest statistical analysis without taking the quantification aspect into account. We introduce a software framework for quantitative analysis of 3D/4D SPECT·CT Tc-99m bone scintigraphy data that incorporates automatic analysis and osseous tissue healthiness classification based on a patient database annotated by an experienced physician. The usage workflow of our framework starts with the input of SPECT and CT data registered on each other. Using CT information we segment osseous tissue also in the co-registered SPECT dataset. Patient weight, injected dose and system calibration factor are read out of the input datasets in order to estimate tracer activity concentration (AC) in kBq/ml and standardized uptake value (SUV). Support vector machine method is used to classify tissue based on the reference patient database that contains quantitative data for normal and abnormal tissue. Evaluation of the framework was done on patient data acquired 3 hours post injection on a SPECT/CT Symbia T6 system using protocol for quantitative estimation of tracer activity. The acquired projection data was reconstructed using Ordered Subset Expectation Maximization with 3D resolution recovery (OSEM-3D) with attenuation and scatter correction. Data analysis showed a significant correlation between AC and bone density measured in Hounsfield Units (HU) for both male and female patients. The proposed methods demonstrate the ability to automatically quantify and distinguish abnormal tissue from healthy one and to increase physicians' diagnostic confidence.