Nauen, David W.

Real-time intraoperative delineation of cancer and non-cancer brain tissues, especially in the eloquent cortex, is critical for thorough cancer resection, lengthening survival, and improving quality of life. Prior studies have established that thresholding optical attenuation values reveals cancer regions with high sensitivity and specificity. However, threshold of a single value disregards local information important to making more robust predictions. Hence, we propose deep convolutional neural networks (CNNs) trained on labeled OCT images and co-occurrence matrix features extracted from these images to synergize attenuation characteristics and texture features. Specifically, we adapt a deep ensemble model trained on 5,831 examples in a training dataset of 7 patients. We obtain 93.31% sensitivity and 97.04% specificity on a holdout set of 4 patients without the need for beam profile normalization using a reference phantom. The segmentation maps produced by parsing the OCT volume and tiling the outputs of our model are in excellent agreement with attenuation mapping-based methods. Our new approach for this important application has considerable implications for clinical translation.

AbstractOligodendroglioma represents a distinctive neoplasm in adults but similar neoplasms occur rarely in children. We studied 20 cases of pediatric oligodendroglioma by SNP array (median age 9 years, range 1–19; 15 grade II and 5 grade III). Cytogenetic abnormalities were present in 8 (53%) grade II and all five anaplastic oligodendrogliomas. Most changes were in the form of deletion and copy neutral loss of heterozygosity (LOH). The most common abnormality was 1p deletion (n = 5). Whole arm 1p19q co‐deletion was present in three cases from adolescent patients and 9p loss in 3, including one low‐grade oligodendroglioma with CDKN2A homozygous deletion. Common losses were largely limited to the anaplastic subset (n = 5) and included 3q29 (n = 3), 11p (n = 3), 17q (n = 3), 4q (n = 2), 6p (n = 2), 13q (n = 2), 14q (n = 2), 17p (n = 2) and whole Ch 18 loss (n = 2). Gains were non‐recurrent except for whole Ch 7 (n = 2) and gain on 12q (n = 2) including the MDM2 locus. Possible germ line LOH (or uniparental disomy) was present in seven cases (35%), with one focal abnormality (22q13.1‐13.2) in two. BRAF‐KIAA1549 fusions and BRAF p.V600E mutations were absent (n = 13 and 8). In summary, cytogenetic alterations in pediatric oligodendrogliomas are characterized mostly by genomic losses, particularly in anaplastic tumors.