Development of Open-Source Toolboxes for Quantifying Spect Images with Fully Automatic Segmentation and Advanced Parameter Extraction Capabilities

Single-Photon Emission Computed Tomography (SPECT) remains to be a popular modality in clinical applications, especially in neuro-degenerative and cardiovascular diseases. Due to the limitation in spatial resolution, SPECT images are often more difficult to read and the diagnostic accuracy depends highly on the reader's experience. Quantitative analysis may provide clinicians with a numerical, objective and reliable source of information to improve the diagnostic accuracy and to evaluate the disease progression. In this proposal, we aim to develop an open-source and free software package that emphasizes on the fully automatic segmentation and advanced parameter extraction to quantify the SPECT images, particularly in neural and cardiac applications. For neural images, we will mainly focus on the [99mTc]TRODAT-SPECT images for Parkinson's Disease management. An SPM-based program will be developed to spatially normalize TRODAT images in a fully automatic way in order to calculate the striatum-to-reference ratio (SRR) of TRODAT uptake. The diagnostic performance of the quantified parameter will be evaluated with a large clinical cohort (n＞400) through a retrospective study. For cardiac images, we will focus on the Thallium-201 images for ischemic heart diseases. A similar SPM-based method will be used to register and segment the Thallium-201 scans to extract the myocardial volume of interest (VOI). Once the myocardial VOI is determined, we will use texture analysis methods developed in our group to quantify the heterogeneity of cardiac uptake of Thallium-201, as a way to predict the degree of coronary stenosis. Again, a large cohort with more than two hundred subjects will be utilized to evaluate the diagnostic performance of our method in predicting the severity of myocardial ischemia. In this proposal, we will be sharing the developed functionalities through royalty-free, open-source software distribution as well as providing our users with technical support and assistance in implementing new functions. This approach may help accelerate the evaluation of the clinical effectiveness of the newly developed methods in this proposal. Furthermore, it possesses a high potential for promoting the academic visibility of the imaging research communities of Taiwan and their international collaboration.

Project ID:PB10408-5719
External Project ID:MOST104-2221-E006-271-MY2