Room: Exhibit Hall
Purpose: Radiomics is defined as the conversion of images to higher dimensional data and the subsequent mining of these data for improved decision support. The primary purpose of this study was to evaluate impact of image pre-processing on radiomics features prediction power in recurrence glioblastoma patients.
Methods: 85 patients data with recurrent and progress GBM who underwent brain MRI were used in this study. All images were preprocessed by different bin width (16, 32, 64, 128 and 256) and also with wavelet transformation method. Following delineation and segmentation of lesions, 860 quantitative 3D features and texture based on intensity histograms (IH), gray level run-length (GLRLM), gray level co-occurrence (GLCM), gray level size-zone texture matrices (GLSZM), neighborhood-difference matrices (NDM), and geometric features were extracted from the 3D-tumor volumes of each lesion. Supervised back-propagation artificial neural network classifier used to predict recurrence glioblastoma patient. Sixty-six percent of data was recruited as test and remained was used as training set.
Results: By using area under ROC curve (AUC) as an assessment index, bin width of 16, 32, 64, 128, 256 and multilayer perceptron (MLP) yield 0.607, 0.714, 0.642, 0.712, 0.642 and 0.623 AUC for prediction recurrence in GBM patient respectively.
Conclusion: The main aim of the current essay was to demonstrate the effect of different pre-processing techniques on radiomics perdition power in recurrence glioblastoma patient. The result of current study shows that, image preprocessing can have a strong impact on Radiomics process in prediction processes.
Texture Analysis, Image Processing, Computer Vision