Elimination of tissue attenuation with classification methods on myocardial perfusion imaging (myocardial perfusion scintigraphy)

Coronary angiography is considered as the gold standard for diagnosis of myocardial ischemia (MI) that is caused by coronary artery disease (CAD); but this method has some disadvantages and medical risks that could end up with death. One of the medical imaging ways; imaging myocardial perfusion with SPECT (single photon emission computed tomography) is non-invasive, cost-effective and easy to perform method. However, this method can be insufficient when breast attenuation exists for women and can cause wrong diagnosis. In order to develop and make easier interpretation of this specific instance, a new computer-assisted diagnosis based model have been developed. In order to develop this model regions of interest (ROI) are selected and features belong to these regions are extracted. After that, various supervised classification methods are applied. Among all tried combinations, best classification accuracy has been obtained as 91.358 % with 0,904 AUC (Area Under Curve). Also body mass indexes of patients are computed and patients are divided into three groups according to these indexes. For these groups, higher classification rates and larger areas under curve (AUC) are obtained. . All these results have been validated by 10-fold Cross-Validation (CV). This developed model could help clinicians on the step of decision before asking angiography, for patients arriving on suspicion of MI on conditions considered breast attenuation in myocardial perfusion scintigraphy (MPS).