2019/08/13 by Engin Dikici, Dikici, Engin, John Ryu +13
Medicine · #Brain Metastases and Treatment #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Medical Imaging Techniques and Applications #Radiomics and Machine Learning in Medical Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1908.04701
openalex publication_date 2019/08/13 · openalex created_date 2022/07/19 · openalex updated_date 2026/07/28
Brain Metastases (BM) complicate 20-40% of cancer cases. BM lesions can\npresent as punctate (1 mm) foci, requiring high-precision Magnetic Resonance\nImaging (MRI) in order to prevent inadequate or delayed BM treatment. However,\nBM lesion detection remains challenging partly due to their structural\nsimilarities to normal structures (e.g., vasculature). We propose a\nBM-detection framework using a single-sequence gadolinium-enhanced T1-weighted\n3D MRI dataset. The framework focuses on detection of smaller (< 15 mm) BM\nlesions and consists of: (1) candidate-selection stage, using Laplacian of\nGaussian approach for highlighting parts of a MRI volume holding higher BM\noccurrence probabilities, and (2) detection stage that iteratively processes\ncropped region-of-interest volumes centered by candidates using a custom-built\n3D convolutional neural network ("CropNet"). Data is augmented extensively\nduring training via a pipeline consisting of random gamma correction and\nelastic deformation stages; the framework thereby maintains its invariance for\na plausible range of BM shape and intensity representations. This approach is\ntested using five-fold cross-validation on 217 datasets from 158 patients, with\ntraining and testing groups randomized per patient to eliminate learning bias.\nThe BM database included lesions with a mean diameter of ~5.4 mm and a mean\nvolume of ~160 mm3. For 90% BM-detection sensitivity, the framework produced on\naverage 9.12 false-positive BM detections per patient (standard deviation of\n3.49); for 85% sensitivity, the average number of false-positives declined to\n5.85. Comparative analysis showed that the framework produces comparable\nBM-detection accuracy with the state-of-art approaches validated for\nsignificantly larger lesions.\n