Cardiac sarcoidosis classification with deep convolutional neural network-based features using polar maps

• Published in: Computers in biology and medicine Vol. 104; pp. 81 - 86
• Main Authors: Togo, Ren, Hirata, Kenji, Manabe, Osamu, Ohira, Hiroshi, Tsujino, Ichizo, Magota, Keiichi, Ogawa, Takahiro, Haseyama, Miki, Shiga, Tohru
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.01.2019; Elsevier Limited
• Subjects: 18F-FDG PET; Aged; Algorithms; Artificial neural networks; Automation; Carbohydrates; Cardiac sarcoidosis (CS); Cardiomyopathies - classification; Cardiomyopathies - diagnostic imaging; Classification; Computed tomography; Computer-aided diagnosis; Convolutional neural network (CNN); Deep learning; Fasting; Feature extraction; Feature selection; Female; Heart; Humans; Machine learning; Male; Medical diagnosis; Medical imaging; Metabolism; Methods; Middle Aged; Neural networks; Neural Networks, Computer; Patients; Physiology; Positron emission; Positron Emission Tomography Computed Tomography; Quantitative analysis; Radiology; Retrospective Studies; Sarcoidosis; Sarcoidosis - classification; Sarcoidosis - diagnostic imaging; Semantics; Sensitivity; Tomography; Ventricle; Japan; Deep learning; 18F-FDG PET; Feature selection; Cardiac sarcoidosis (CS); Machine learning; Radiology; Feature extraction; Convolutional neural network (CNN); Computer-aided diagnosis; F-FDG PET
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2018.11.008

The aim of this study was to determine whether deep convolutional neural network (DCNN)-based features can represent the difference between cardiac sarcoidosis (CS) and non-CS using polar maps. A total of 85 patients (33 CS patients and 52 non-CS patients) were analyzed as our study subjects. One radiologist reviewed PET/CT images and defined the left ventricle region for the construction of polar maps. We extracted high-level features from the polar maps through the Inception-v3 network and evaluated their effectiveness by applying them to a CS classification task. Then we introduced the ReliefF algorithm in our method. The standardized uptake value (SUV)-based classification method and the coefficient of variance (CoV)-based classification method were used as comparative methods. Sensitivity, specificity and the harmonic mean of sensitivity and specificity of our method with the ReliefF algorithm were 0.839, 0.870 and 0.854, respectively. Those of the SUVmax-based classification method were 0.468, 0.710 and 0.564, respectively, and those of the CoV-based classification method were 0.655, 0.750 and 0.699, respectively. The DCNN-based high-level features may be more effective than low-level features used in conventional quantitative analysis methods for CS classification. •A method for detection of cardiac sarcoidosis was proposed.•The DCNN-based features were effective for the cardiac sarcoidosis classification.•Feature selection algorithm was effective for improving the performance.