Search

Background and Objective: Treatment approaches for colorectal cancer (CRC) are highly dependent on the molecular subtype, as immunotherapy has shown efficacy in cases with microsatellite instability (MSI) but is ineffective for the microsatellite stable (MSS) subtype. There is promising potential in utilizing deep neural networks (DNNs) to automate the differentiation of CRC subtypes by analyzing hematoxylin and eosin (H&E) stained whole-slide images (WSIs). Due to the extensive size of WSIs, multiple instance learning (MIL) techniques are typically explored. However, existing MIL methods focus on identifying the most representative image patches for classification, which may result in the loss of critical information. Additionally, these methods often overlook clinically relevant information, like the tendency for MSI class tumors to predominantly occur on the proximal (right side) colon., Methods: We introduce 'CIMIL-CRC', a DNN framework that: (1) solves the MSI/MSS MIL problem by efficiently combining a pre-trained feature extraction model with principal component analysis (PCA) to aggregate information from all patches, and (2) integrates clinical priors, particularly the tumor location within the colon, into the model to enhance patient-level classification accuracy. We assessed our CIMIL-CRC method using the average area under the receiver operating characteristic curve (AUROC) from a 5-fold cross-validation experimental setup for model development on the TCGA-CRC-DX cohort, contrasting it with a baseline patch-level classification, a MIL-only approach, and a clinically-informed patch-level classification approach., Results: Our CIMIL-CRC outperformed all methods (AUROC: 0.92±0.002 (95% CI 0.91-0.92), vs. 0.79±0.02 (95% CI 0.76-0.82), 0.86±0.01 (95% CI 0.85-0.88), and 0.87±0.01 (95% CI 0.86-0.88), respectively). The improvement was statistically significant. To the best of our knowledge, this is the best result achieved for MSI/MSS classification on this dataset., Conclusion: Our CIMIL-CRC method holds promise for offering insights into the key representations of histopathological images and suggests a straightforward implementation., Competing Interests: Declaration of competing interest We affirm that there exist no financial or personal affiliations with individuals or organizations that could unjustly sway our research. Further, we have no professional or personal interests, of any nature, in any products, services, and/or companies that could be interpreted as influencing the stance taken in, or the review of, the manuscript under consideration., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Background: Gliomas are a major cause of cancer-related death among children, adolescents, and young adults (age 0-40 years). Primary mismatch repair deficiency (MMRD) is a pan-cancer mechanism with unique biology and therapeutic opportunities. We aimed to determine the extent and impact of primary MMRD in gliomas among children, adolescents, and young adults., Methods: Clinical and molecular data were collected from a population-based cohort of children, adolescents, and young adults with gliomas from Toronto (TOR-Ped, age 0-18 years, collected Jan 1, 2000, to Dec 31, 2021; and TOR-AYA, age 18-40 years, collected Jan 1, 2000, to June 30, 2019). Additional validation paediatric cohorts from St Jude Children's Research Hospital (0-18 years, 2015-21) and the Children's Brain Tumor Network (0-18 years, 1981-2021) were used. Functional genomic tools were applied with the primary aim of assessing primary MMRD prevalence among glioma subgroups and germline impact. To evaluate the effect of primary MMRD on therapy and overall survival, Kaplan-Meier estimates were used on an additional cohort of patients with primary MMRD gliomas treated with immunotherapy., Findings: 1389 gliomas were included in the study. The prevalence of primary MMRD ranged between 3·7% and 12·4% in high-grade gliomas (overall 30 of 483; 6·2%, 95% CI 4·2-8·7) and less than 1% in low-grade gliomas (four of 899; 0·4%, 0·1-1·1; p<0·0001 by χ 2 test). Specific molecular analysis for all gliomas showed that primary MMRD was absent among oligodendrogliomas (none of 67) and uncommon in BRAF V600E gliomas (one of 110) and histone mutant-driven gliomas (one of 150). In the paediatric age group (<18 years), primary MMRD was common in IDH WT and H3 WT gliomas harbouring pathogenic TP53 variants (21 of 61; 34·4%, 22·7-47·7) and in malignant IDH mut gliomas (five of eight; 62·5%, 24·5-91·5). Germline aetiology accounted for 33 (94·3%) of 35 primary MMRD gliomas, including children, adolescents, and young adults with previously unrecognised Lynch syndrome. Survival was poor for patients with primary MMRD gliomas. Particularly poor survival was observed for those with IDH mut astrocytomas with primary MMRD when compared with those with mismatch repair-proficient gliomas (HR 12·6, 95% CI 2·8-57·5; p=0·0011 by multivariable Cox regression). Immune checkpoint blockade was associated with improved survival for patients with primary MMRD gliomas compared with conventional chemoradiotherapy regimens (HR 0·4, 0·3-0·7; p=0·0017 by multivariable Cox regression), regardless of age or germline status., Interpretation: Primary MMRD is more common than previously reported in gliomas in children, adolescents, and young adults, is enriched in specific molecular subgroups, and is associated with poor outcomes. Accurate detection, genetic testing, early diagnosis through surveillance, and implementation of immunotherapy might improve survival for these patients., Funding: The Canadian Institutes for Health Research, Stand Up to Cancer-Bristol Myers Squibb Catalyst, US National Institutes of Health, Canadian Cancer Society, Brain Canada, The V Foundation for Cancer Research, BioCanRx, Canada's Immunotherapy Network, Harry and Agnieszka Hall, Meagan's Hug, BRAINchild Canada, and the LivWise Foundation., Competing Interests: Declaration of interests AFG reports consulting fees and speaker honoraria from Servier Canada. DST reports consulting fees for Need (getneed.com). GG receives research funds from IBM, Pharmacyclics/AbbVie, Bayer, Genentech, Calico, and Ultima Genomics. GG is an inventor on patent applications related to MSMuTect, MSMutSig, POLYSOLVER, SignatureAnalyzer-GPU, MSIDetect/MSEye, LOGIC, and MinimuMM-seq; is a founder, consultant, and has privately held equity in Scorpion Therapeutics; is a founder of and has privately held equity in PreDICTA Biosciences; and has consulted for Merck. JK reports funding from Canadian Institutes of Health Research and Brain Canada; consulting as a neuropathologist for the Ontario Forensic Pathology Services; provides expert witness testimony, support to attend Canadian and American Association of Neuropathologists annual meetings, and support from the Department of Laboratory Medicine and Molecular Biology at Sunnybrook Health Science Centre to undertake a forensic neuropathology sabbatical; is the Vice Chair of the Royal College of Physicians and Surgeons of Canada Specialty Committee in Neuropathology; and reports leadership positions in the Canadian Association of Neuropathologists. SDa reports grants from Ontario Institute for Cancer Research, New Frontiers in Research Fund, Canadian Institutes of Health Research, Canadian Cancer Society, Breast Cancer Canada, American Brain Tumor Association, Synaptive Medical, VPIX Medical, Gratitude 10, and Macbeth Family Trust; is the Keenan Chair in Surgery at the Temerty Faculty of Medicine at the University of Toronto; reports royalties or licenses from Oxford University Press; and holds leaderships roles at Ontario Health (Cancer Care Ontario) and Subortical Surgery Group. All other authors declare no competing interests., (Copyright © 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.)