2019/07/25 by Jason G. Parker, Parker, Jason G, PhD +17
Medicine · #Data Analysis #FOS: Physical sciences #Glioma Diagnosis and Treatment #MRI in cancer diagnosis #Medical Physics (physics.med-ph) #Radiomics and Machine Learning in Medical Imaging #Statistics and Probability (physics.data-an)
paper · pdf · doi:10.48550/arxiv.1907.11161
openalex publication_date 2019/07/25 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
We propose a statistical multiscale mapping approach to identify microscopic\nand molecular heterogeneity across a tumor microenvironment using\nmultiparametric MR (mp-MR). Twenty-nine patients underwent pre-surgical mp-MR\nfollowed by MR-guided stereotactic core biopsy. The locations of the biopsy\ncores were identified in the pre-surgical images using stereotactic bitmaps\nacquired during surgery. Feature matrices mapped the multiparametric voxel\nvalues in the vicinity of the biopsy cores to the pathologic outcome variables\nfor each patient and logistic regression tested the individual and collective\npredictive power of the MR contrasts. A non-parametric weighted k-nearest\nneighbor classifier evaluated the feature matrices in a leave-one-out cross\nvalidation design across patients. Resulting class membership probabilities\nwere converted to chi-square statistics to develop full-brain parametric maps,\nimplementing Gaussian random field theory to estimate inter-voxel dependencies.\nCorrections for family-wise error rates were performed using Benjamini-Hochberg\nand random field theory, and the resulting accuracies were compared. The\ncombination of all five image contrasts correlated with outcome (P<.001) for\nall four microscopic variables. The probabilistic mapping method using\nBenjamini-Hochberg generated statistically significant results (P<.05) for\nthree of the four dependent variables: 1) IDH1, 2) MGMT, and 3) microvascular\nproliferation, with an average classification accuracy of 0.984 +/- 0.02 and an\naverage classification sensitivity of 1.567% +/- 0.967. The images corrected by\nrandom field theory demonstrated improved classification accuracy (0.989 +/-\n0.008) and classification sensitivity (5.967% +/- 2.857) compared with\nBenjamini-Hochberg. Microscopic and molecular tumor properties can be assessed\nwith statistical confidence across the brain from minimally-invasive, mp-MR.\n