A DATASET OF SPECT MPI FOR DEVELOPING AND EVALUATING MACHINE LEARNING MODELS FOR MYOCARDIAL LESION DETECTION

128 views

Abstract

Myocardial perfusion imaging (MPI) with a single photon emission tomography (SPECT) is a potential research area for machine learning for diagnostic support. However, the research results on this issue are few and difficult to apply and develop because of the lack of a standard database. Most detection algorithms for detecting myocardial lesion with SPECT MPI image analysis are based on private data sets or are published with unknown information. This makes it difficult to compare performance between different methods and development. In order to solve this problem, we bring the standard data set to SPECT MPI to evaluate the methods in detecting myocardial lesion. In addition, we present machine learning models that are currently in common use and compare their performance based on this data set. Models are designed to serve as a basis for comparison with later development models. We believe that the database and the basic model will contribute to the development of applied machine learning studies in supporting SPECT MPI diagnostics.