2019/10/04 by Markus Frey, M. M. Frey, Matthias Nau +2
Biochemistry, Genetics and Molecular Biology · Computer Science · Engineering · Medicine · Neuroscience · #Advanced Neural Network Applications #Artificial intelligence #Autoencoder #Brain Tumor Detection and Classification #Brain tumor #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Computer vision #Convolutional neural network #Deep learning #FOS: Biological sciences #FOS: Computer and information sciences #FOS: Electrical engineering #Fluid-attenuated inversion recovery #Image and Video Processing (eess.IV) #Magnetic resonance imaging #Medicine #Neurons and Cognition (q-bio.NC) #Pathology #Pattern recognition (psychology) #Preprocessor #Radiology #Radiomics and Machine Learning in Medical Imaging #Segmentation #cs.CV #eess.IV #electronic engineering #information engineering #q-bio.NC
paper · pdf · doi:10.48550/arxiv.1910.02058
published in arXiv (Cornell University) (Cornell University)
arxiv created 2019/10/04 · openalex publication_date 2019/10/04 · arxiv updated 2019/10/07 · openalex created_date 2022/07/28 · openalex updated_date 2026/08/05
Early diagnosis and accurate segmentation of brain tumors are imperative for\nsuccessful treatment. Unfortunately, manual segmentation is time consuming,\ncostly and despite extensive human expertise often inaccurate. Here, we present\nan MRI-based tumor segmentation framework using an autoencoder-regularized\n3D-convolutional neural network. We trained the model on manually segmented\nstructural T1, T1ce, T2, and Flair MRI images of 335 patients with tumors of\nvariable severity, size and location. We then tested the model using\nindependent data of 125 patients and successfully segmented brain tumors into\nthree subregions: the tumor core (TC), the enhancing tumor (ET) and the whole\ntumor (WT). We also explored several data augmentations and preprocessing steps\nto improve segmentation performance. Importantly, our model was implemented on\na single NVIDIA GTX1060 graphics unit and hence optimizes tumor segmentation\nfor widely affordable hardware. In sum, we present a memory-efficient and\naffordable solution to tumor segmentation to support the accurate diagnostics\nof oncological brain pathologies.\n