Magnetic resonance imaging radiomics analysis for predicting hepatocellular carcinoma
NSM Haniff, MKBA Karim, NS Ali… - 2021 International …, 2021 - ieeexplore.ieee.org
2021 International Congress of Advanced Technology and Engineering …, 2021•ieeexplore.ieee.org
Current technology allows for more accurate and precise diagnosis that able to classify the
tumour staging by quantifying the features extraction and medical images analysis. Hence,
this study aimed to evaluate radiomic features of Hepatocellular Carcinoma (HCC) based on
magnetic resonance imaging (MRI) modality and to classify the tumour based on their
staging. The image data were filtered by Laplacian Sharpening Images, and segmented by
semi-automatic segmentation. Features such as shape features, first-order and second order …
tumour staging by quantifying the features extraction and medical images analysis. Hence,
this study aimed to evaluate radiomic features of Hepatocellular Carcinoma (HCC) based on
magnetic resonance imaging (MRI) modality and to classify the tumour based on their
staging. The image data were filtered by Laplacian Sharpening Images, and segmented by
semi-automatic segmentation. Features such as shape features, first-order and second order …
Current technology allows for more accurate and precise diagnosis that able to classify the tumour staging by quantifying the features extraction and medical images analysis. Hence, this study aimed to evaluate radiomic features of Hepatocellular Carcinoma (HCC) based on magnetic resonance imaging (MRI) modality and to classify the tumour based on their staging. The image data were filtered by Laplacian Sharpening Images, and segmented by semi-automatic segmentation. Features such as shape features, first-order and second order statistic (GLSZM) were extracted from the segmented images. 51 data patients retrieved from The Cancer Imaging Archive (TCIA) browser and total forty-eight radiomic features were extracted from each patient. Features extracted were categorized into two group of HCC; group one (stage I and II) and group two (stage III and IV) and classified using automated machine learning (AutoML). Tree Based Pipeline Optimization Tool (TPOT) algorithm is chosen to perform the classification. TPOT algorithm selected the best pipeline with the highest accuracy and its performance was assessed using several performance metrics. The features range was seen to be overlapped between original images and images with enhancement after undergoes normalization. Decision Tree (DT) classifier is chosen as the best pipeline among TPOT algorithm for the features extracted from enhanced images. The accuracy, precision and recall for the pipeline are 0.846, 0.75 and 0.75. Area under ROC curve for DT classifier is 0.917. This study shows significance of image enhancement in pre-processing steps and AutoML.
ieeexplore.ieee.org
以上显示的是最相近的搜索结果。 查看全部搜索结果