scholarly journals A Deep Learning Approach for Rapid Mutational Screening in Melanoma

2019 ◽  
Author(s):  
Randie H. Kim ◽  
Sofia Nomikou ◽  
Nicolas Coudray ◽  
George Jour ◽  
Zarmeena Dawood ◽  
...  
Keyword(s):  
Deep Learning ◽  
Structural Characteristics ◽  
Tumor Biology ◽  
Model Performance ◽  
Area Under The Curve ◽  
The Cancer Genome Atlas ◽  
Clinical Settings ◽  
Mutational Screening ◽  
Cancer Genome Atlas ◽  
Whole Slide Images

AbstractImage-based analysis as a rapid method for mutation detection can be advantageous in research or clinical settings when tumor tissue is limited or unavailable for direct testing. Here, we applied a deep convolutional neural network (CNN) to whole slide images of melanomas from 256 patients and developed a fully automated model that first selects for tumor-rich areas (Area Under the Curve AUC=0.96) then predicts for the presence of mutated BRAF in our test set (AUC=0.72) Model performance was cross-validated on melanoma images from The Cancer Genome Atlas (AUC=0.75). We confirm that the mutated BRAF genotype is linked to phenotypic alterations at the level of the nucleus through saliency mapping and pathomics analysis, which reveal that cells with mutated BRAF exhibit larger and rounder nuclei. Not only do these findings provide additional insights on how BRAF mutations affects tumor structural characteristics, deep learning-based analysis of histopathology images have the potential to be integrated into higher order models for understanding tumor biology, developing biomarkers, and predicting clinical outcomes.

scholarly journals Mining the plasma-proteome associated genes in patients with gastro-esophageal cancers for biomarker discovery

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Frederick S. Vizeacoumar ◽  
Hongyu Guo ◽  
Lynn Dwernychuk ◽  
Adnan Zaidi ◽  
Andrew Freywald ◽  
...  
Keyword(s):  
Biomarker Discovery ◽  
Predictive Accuracy ◽  
Area Under The Curve ◽  
Cancer Recurrence ◽  
The Cancer Genome Atlas ◽  
Svm Classifier ◽  
Plasma Proteome ◽  
Cancer Genome Atlas ◽  
Novel Biomarkers ◽  
Evaluation Metric

AbstractGastro-esophageal (GE) cancers are one of the major causes of cancer-related death in the world. There is a need for novel biomarkers in the management of GE cancers, to yield predictive response to the available therapies. Our study aims to identify leading genes that are differentially regulated in patients with these cancers. We explored the expression data for those genes whose protein products can be detected in the plasma using the Cancer Genome Atlas to identify leading genes that are differentially regulated in patients with GE cancers. Our work predicted several candidates as potential biomarkers for distinct stages of GE cancers, including previously identified CST1, INHBA, STMN1, whose expression correlated with cancer recurrence, or resistance to adjuvant therapies or surgery. To define the predictive accuracy of these genes as possible biomarkers, we constructed a co-expression network and performed complex network analysis to measure the importance of the genes in terms of a ratio of closeness centrality (RCC). Furthermore, to measure the significance of these differentially regulated genes, we constructed an SVM classifier using machine learning approach and verified these genes by using receiver operator characteristic (ROC) curve as an evaluation metric. The area under the curve measure was > 0.9 for both the overexpressed and downregulated genes suggesting the potential use and reliability of these candidates as biomarkers. In summary, we identified leading differentially expressed genes in GE cancers that can be detected in the plasma proteome. These genes have potential to become diagnostic and therapeutic biomarkers for early detection of cancer, recurrence following surgery and for development of targeted treatment.

scholarly journals Deep Learning for Automatic Subclassification of Gastric Carcinoma Using Whole-Slide Histopathology Images

Cancers ◽  
2021 ◽  
Vol 13 (15) ◽  
pp. 3811
Author(s):  
Hyun-Jong Jang ◽  
In-Hye Song ◽  
Sung-Hak Lee
Keyword(s):  
Deep Learning ◽  
Relative Proportion ◽  
The Cancer Genome Atlas ◽  
Cancer Tissue ◽  
Mucinous Tumor ◽  
Receiver Operating Characteristic Curves ◽  
Tumor Tissues ◽  
Cancer Genome Atlas ◽  
Stomach Adenocarcinoma ◽  
Tumor Types

Histomorphologic types of gastric cancer (GC) have significant prognostic values that should be considered during treatment planning. Because the thorough quantitative review of a tissue slide is a laborious task for pathologists, deep learning (DL) can be a useful tool to support pathologic workflow. In the present study, a fully automated approach was applied to distinguish differentiated/undifferentiated and non-mucinous/mucinous tumor types in GC tissue whole-slide images from The Cancer Genome Atlas (TCGA) stomach adenocarcinoma dataset (TCGA-STAD). By classifying small patches of tissue images into differentiated/undifferentiated and non-mucinous/mucinous tumor tissues, the relative proportion of GC tissue subtypes can be easily quantified. Furthermore, the distribution of different tissue subtypes can be clearly visualized. The patch-level areas under the curves for the receiver operating characteristic curves for the differentiated/undifferentiated and non-mucinous/mucinous classifiers were 0.932 and 0.979, respectively. We also validated the classifiers on our own GC datasets and confirmed that the generalizability of the classifiers is excellent. The results indicate that the DL-based tissue classifier could be a useful tool for the quantitative analysis of cancer tissue slides. By combining DL-based classifiers for various molecular and morphologic variations in tissue slides, the heterogeneity of tumor tissues can be unveiled more efficiently.

scholarly journals A Novel m6A Gene Signature Associated With Regulatory Immune Function for Prognosis Prediction in Clear-Cell Renal Cell Carcinoma

2021 ◽  
Vol 8 ◽  
Author(s):  
Siteng Chen ◽  
Ning Zhang ◽  
Encheng Zhang ◽  
Tao Wang ◽  
Liren Jiang ◽  
...  
Keyword(s):  
Risk Score ◽  
Clear Cell ◽  
Area Under The Curve ◽  
The Cancer Genome Atlas ◽  
Survival Prediction ◽  
Cell Renal Cell Carcinoma ◽  
Training Cohort ◽  
Rna Methylation ◽  
Cancer Genome Atlas ◽  
Genome Atlas

The important role of N6-methyladenosine (m6A) RNA methylation regulator in carcinogenesis and progression of clear-cell renal cell carcinoma (ccRCC) is poorly understood by now. In this study, we performed comprehensive analyses of m6A RNA methylation regulators in 975 ccRCC samples and 332 adjacent normal tissues and identified ccRCC-related m6A regulators. Moreover, the m6A diagnostic score based on ccRCC-related m6A regulators could accurately distinguish ccRCC from normal tissue in the Meta-cohort, which was further validated in the independent GSE-cohort and The Cancer Genome Atlas-cohort, with an area under the curve of 0.924, 0.867, and 0.795, respectively. Effective survival prediction of ccRCC by m6A risk score was also identified in the Cancer Genome Atlas training cohort and verified in the testing cohort and the independent GSE22541 cohort, with hazard ratio values of 3.474, 1.679, and 2.101 in the survival prognosis, respectively. The m6A risk score was identified as a risk factor of overall survival in ccRCC patients by the univariate Cox regression analysis, which was further verified in both the training cohort and the independent validation cohort. The integrated nomogram combining m6A risk score and predictable clinicopathologic factors could accurately predict the survival status of the ccRCC patients, with an area under the curve values of 85.2, 82.4, and 78.3% for the overall survival prediction in 1-, 3- and 5-year, respectively. Weighted gene co-expression network analysis with functional enrichment analysis indicated that m6A RNA methylation might affect clinical prognosis through regulating immune functions in patients with ccRCC.

scholarly journals Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists

BMJ Open ◽  
2020 ◽  
Vol 10 (9) ◽  
pp. e036423
Author(s):  
Zhigang Song ◽  
Chunkai Yu ◽  
Shuangmei Zou ◽  
Wenmiao Wang ◽  
Yong Huang ◽  
...  
Keyword(s):  
Deep Learning ◽  
Colorectal Adenoma ◽  
Detection System ◽  
Model Performance ◽  
Area Under The Curve ◽  
Learning Model ◽  
Feature Maps ◽  
Adenoma Detection ◽  
Microscopic Evaluation ◽  
Deep Learning Model

ObjectivesThe microscopic evaluation of slides has been gradually moving towards all digital in recent years, leading to the possibility for computer-aided diagnosis. It is worthwhile to know the similarities between deep learning models and pathologists before we put them into practical scenarios. The simple criteria of colorectal adenoma diagnosis make it to be a perfect testbed for this study.DesignThe deep learning model was trained by 177 accurately labelled training slides (156 with adenoma). The detailed labelling was performed on a self-developed annotation system based on iPad. We built the model based on DeepLab v2 with ResNet-34. The model performance was tested on 194 test slides and compared with five pathologists. Furthermore, the generalisation ability of the learning model was tested by extra 168 slides (111 with adenoma) collected from two other hospitals.ResultsThe deep learning model achieved an area under the curve of 0.92 and obtained a slide-level accuracy of over 90% on slides from two other hospitals. The performance was on par with the performance of experienced pathologists, exceeding the average pathologist. By investigating the feature maps and cases misdiagnosed by the model, we found the concordance of thinking process in diagnosis between the deep learning model and pathologists.ConclusionsThe deep learning model for colorectal adenoma diagnosis is quite similar to pathologists. It is on-par with pathologists’ performance, makes similar mistakes and learns rational reasoning logics. Meanwhile, it obtains high accuracy on slides collected from different hospitals with significant staining configuration variations.

scholarly journals The Impact of Digital Histopathology Batch Effect on Deep Learning Model Accuracy and Bias

2020 ◽  
Author(s):  
Frederick M. Howard ◽  
James Dolezal ◽  
Sara Kochanny ◽  
Jefree Schulte ◽  
Heather Chen ◽  
...  
Keyword(s):  
Deep Learning ◽  
Expression Patterns ◽  
Model Performance ◽  
Batch Effect ◽  
The Cancer Genome Atlas ◽  
Driver Mutations ◽  
Cancer Subtypes ◽  
Breast Cancer Cohort ◽  
Quadratic Programming Method ◽  
The Impact

AbstractThe Cancer Genome Atlas (TCGA) is one of the largest biorepositories of digital histology. Deep learning (DL) models have been trained on TCGA to predict numerous features directly from histology, including survival, gene expression patterns, and driver mutations. However, we demonstrate that these features vary substantially across tissue submitting sites in TCGA for over 3,000 patients with six cancer subtypes. Additionally, we show that histologic image differences between submitting sites can easily be identified with DL. This site detection remains possible despite commonly used color normalization and augmentation methods, and we quantify the digital image characteristics constituting this histologic batch effect. As an example, we show that patient ethnicity within the TCGA breast cancer cohort can be inferred from histology due to site-level batch effect, which must be accounted for to ensure equitable application of DL. Batch effect also leads to overoptimistic estimates of model performance, and we propose a quadratic programming method to guide validation that abrogates this bias.

scholarly journals A Six-lncRNA Signature for Immunophenotype Prediction of Glioblastoma Multiforme

2021 ◽  
Vol 11 ◽  
Author(s):  
Ming Gao ◽  
Xinzhuang Wang ◽  
Dayong Han ◽  
Enzhou Lu ◽  
Jian Zhang ◽  
...  
Keyword(s):  
Glioblastoma Multiforme ◽  
Immune Cell ◽  
Area Under The Curve ◽  
Principal Component ◽  
Enrichment Analysis ◽  
Gene Set Enrichment Analysis ◽  
The Cancer Genome Atlas ◽  
Non Coding Rna ◽  
Cancer Genome Atlas ◽  
The Central Nervous System

Glioblastoma multiforme (GBM) is the most aggressive primary tumor of the central nervous system. As biomedicine advances, the researcher has found the development of GBM is closely related to immunity. In this study, we evaluated the GBM tumor immunoreactivity and defined the Immune-High (IH) and Immune-Low (IL) immunophenotypes using transcriptome data from 144 tumors profiled by The Cancer Genome Atlas (TCGA) project based on the single-sample gene set enrichment analysis (ssGSEA) of five immune expression signatures (IFN-γ response, macrophages, lymphocyte infiltration, TGF-β response, and wound healing). Next, we identified six immunophenotype-related long non-coding RNA biomarkers (im-lncRNAs, USP30-AS1, HCP5, PSMB8-AS1, AL133264.2, LINC01684, and LINC01506) by employing a machine learning computational framework combining minimum redundancy maximum relevance algorithm (mRMR) and random forest model. Moreover, the expression level of identified im-lncRNAs was converted into an im-lncScore using the normalized principal component analysis. The im-lncScore showed a promising performance for distinguishing the GBM immunophenotypes with an area under the curve (AUC) of 0.928. Furthermore, the im-lncRNAs were also closely associated with the levels of tumor immune cell infiltration in GBM. In summary, the im-lncRNA signature had important clinical implications for tumor immunophenotyping and guiding immunotherapy in glioblastoma patients in future.

scholarly journals Spatial heterogeneity and organization of tumor mutation burden and immune infiltrates within tumors based on whole slide images correlated with patient survival in bladder cancer

2019 ◽  
Author(s):  
Hongming Xu ◽  
Sunho Park ◽  
Jean René Clemenceau ◽  
Jinhwan Choi ◽  
Nathan Radakovich ◽  
...  
Keyword(s):  
Bladder Cancer ◽  
Spatial Heterogeneity ◽  
Tumor Cells ◽  
Tumor Infiltrating Lymphocytes ◽  
The Cancer Genome Atlas ◽  
Mutation Burden ◽  
Prognostic Utility ◽  
Cancer Genome Atlas ◽  
Infiltrating Lymphocytes ◽  
Whole Slide Images

AbstractHigh-TMB (TMB-H) could result in an increased number of neoepitopes from somatic mutations expressed by a patient’s own tumor cell which can be recognized and targeted by neighboring tumor-infiltrating lymphocytes (TILs). Deeper understanding of spatial heterogeneity and organization of tumor cells and their neighboring immune infiltrates within tumors could provide new insights into tumor progression and treatment response. Here we developed and applied computational approaches using digital whole slide images (WSIs) to investigate spatial heterogeneity and organization of regions harboring TMB-H tumor cells and TILs within tumors, and its prognostic utility. In experiments using WSIs from The Cancer Genome Atlas bladder cancer (BLCA), our findings show that WSI-based approaches can reliably predict patient-level TMB status and delineate spatial TMB heterogeneity and co-organization with TILs. TMB-H patients with low spatial heterogeneity enriched with high TILs show improved overall survival indicating a prognostic role of spatial TMB and TILs information in BLCA.

scholarly journals Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

2021 ◽  
Vol 12 (1) ◽  
Author(s):  
Gang Yu ◽  
Kai Sun ◽  
Chao Xu ◽  
Xing-Hua Shi ◽  
Chong Wu ◽  
...  
Keyword(s):  
Artificial Intelligence ◽  
Colorectal Cancer ◽  
Deep Learning ◽  
Supervised Learning ◽  
Area Under The Curve ◽  
Patient Level ◽  
Significant Difference ◽  
The Mean ◽  
Whole Slide Images ◽  
Better Than

AbstractMachine-assisted pathological recognition has been focused on supervised learning (SL) that suffers from a significant annotation bottleneck. We propose a semi-supervised learning (SSL) method based on the mean teacher architecture using 13,111 whole slide images of colorectal cancer from 8803 subjects from 13 independent centers. SSL (~3150 labeled, ~40,950 unlabeled; ~6300 labeled, ~37,800 unlabeled patches) performs significantly better than the SL. No significant difference is found between SSL (~6300 labeled, ~37,800 unlabeled) and SL (~44,100 labeled) at patch-level diagnoses (area under the curve (AUC): 0.980 ± 0.014 vs. 0.987 ± 0.008, P value = 0.134) and patient-level diagnoses (AUC: 0.974 ± 0.013 vs. 0.980 ± 0.010, P value = 0.117), which is close to human pathologists (average AUC: 0.969). The evaluation on 15,000 lung and 294,912 lymph node images also confirm SSL can achieve similar performance as that of SL with massive annotations. SSL dramatically reduces the annotations, which has great potential to effectively build expert-level pathological artificial intelligence platforms in practice.

Systematic analysis of parameters predicting pathological axillary status (ypN0 vs. ypN+) in patients converting from cN+ to ycN0 through primary systemic therapy (PST).

2017 ◽  
Vol 35 (15_suppl) ◽  
pp. 560-560
Author(s):  
Cornelia Liedtke ◽  
Hans-Christian Kolberg ◽  
Laura Kerschke ◽  
Dennis Goerlich ◽  
Ingo Bauerfeind ◽  
...  
Keyword(s):  
Multivariate Analysis ◽  
Tumor Biology ◽  
Model Performance ◽  
Area Under The Curve ◽  
Axillary Staging ◽  
Patient Counseling ◽  
Primary Systemic Therapy ◽  
Systematic Analysis ◽  
Logistic Regression Models ◽  
Leave One Out

560 Background: Optimization of axillary staging in patients converting from cN+ to ycN0 through PST is needed. The aim of this analysis was to develop a nomogram predicting the probability of ypN+ after PST based on clinical/pathological parameters. Methods: Patients converting from cN+ to ycN0 through PST from a prospective study (SENTINA arm C) were included. Univariate/multivariate analyses were carried out for 14 clinical/pathological parameters to predict ypN+ using logistic regression models. Odds ratios and 95% confidence intervals were reported. Model performance was assessed by leave-one-out cross-validation (LOOCV at .5 cut-offs) and ROC analyses. Calculations were performed using the SAS Software (Version 9.4). Results: 553 patients were assessed. Stepwise backward variable selection based on a multivariate analysis of all significant parameters resulted in a model (5M, Table, N = 369 evaluable) including ER (3.81; 2.25-6.44), multifocality (2.22; 1.26-3.92), LVI (9.16; 4.68-17.90), detection of SLN after PST (.50; .26-.95) and ycT (1.03; 1.01-1.06). In LOOCV, this model had an area under the curve of .81. Multivariate analysis of parameters available preoperatively showed an association between ypN0/ypN+, ER and ycT. Full subset selection resulted in a model (2M, N = 414) containing only ER (4.36; 2.80, 6.81) and ycT (1.04; 1.02, 1.07). Conclusions: A prediction model including parameters evaluable before/after definitive surgery resulted in a nomogram with acceptable accuracy. Limitation to parameters evaluable before surgery (i.e. ER, ycT) showed reduced accuracy that was comparable/superior to accuracy of using individual parameters. Since tumor biology was the strongest parameter in our models, we hypothesize that modern tumor biologic parameters such as gene expression profiling might optimize prediction of axillary status after PST improving patient counseling. [Table: see text]

scholarly journals Detection of Lung Nodules in Micro-CT Imaging Using Deep Learning

Tomography ◽  
2021 ◽  
Vol 7 (3) ◽  
pp. 358-372
Author(s):  
Matthew D. Holbrook ◽  
Darin P. Clark ◽  
Rutulkumar Patel ◽  
Yi Qi ◽  
Alex M. Bassil ◽  
...  
Keyword(s):  
Deep Learning ◽  
Lung Nodule ◽  
Model Performance ◽  
Area Under The Curve ◽  
Simulated Data ◽  
Real Data ◽  
Lung Tumors ◽  
Training Data ◽  
Micro Ct ◽  
Micro Computed Tomography

We are developing imaging methods for a co-clinical trial investigating synergy between immunotherapy and radiotherapy. We perform longitudinal micro-computed tomography (micro-CT) of mice to detect lung metastasis after treatment. This work explores deep learning (DL) as a fast approach for automated lung nodule detection. We used data from control mice both with and without primary lung tumors. To augment the number of training sets, we have simulated data using real augmented tumors inserted into micro-CT scans. We employed a convolutional neural network (CNN), trained with four competing types of training data: (1) simulated only, (2) real only, (3) simulated and real, and (4) pretraining on simulated followed with real data. We evaluated our model performance using precision and recall curves, as well as receiver operating curves (ROC) and their area under the curve (AUC). The AUC appears to be almost identical (0.76–0.77) for all four cases. However, the combination of real and synthetic data was shown to improve precision by 8%. Smaller tumors have lower rates of detection than larger ones, with networks trained on real data showing better performance. Our work suggests that DL is a promising approach for fast and relatively accurate detection of lung tumors in mice.

Sign in / Sign up

Export Citation Format

Share Document