Determination of biologic and prognostic feature scores from whole slide histology images using deep learning.

2020 ◽  
Vol 38 (15_suppl) ◽  
pp. e17527-e17527 ◽  
Author(s):  
Okyaz Eminaga ◽  
Mahmoud Abbas ◽  
Axel Semjonow ◽  
James D Brooks ◽  
Daniel Rubin
Keyword(s):  
Deep Learning ◽  
Androgen Receptor ◽  
Regression Model ◽  
Cox Regression ◽  
Risk Group ◽  
Receptor Protein ◽  
Gleason Grade ◽  
Cox Regression Model ◽  
Survival Analyses ◽  
Histopathological Feature

e17527 Background: In cancer, histopathology is a reflection of the underlying molecular changes in the cancer cells and provides prognostic information on the risk of disease progression. Therefore, whole slide images may harbor histopathological features that have a biological association and are prognostic. Methods: This study has extracted histopathological feature scores generated from hematoxylin and eosin (HE) histology images based on deep learning models developed for the detection of pathological findings related to prostate cancer (PCa). Correlation analyses between the histopathological feature scores and the most relevant genomic alterations related to PCa were performed based on the original results and diagnostic histology images from TCGA PRAD study (n = 251). We extracted feature scores from tumor lesions after applying tumor segmentation and several data transformation using five models developed for detection of cribriform or ductal morphologies, Gleason patterns 3 and 4, and the presumed tumor precursor. For prognostic evaluation, we performed survival analyses of 371 patients from the TCGA PRAD dataset with biochemical recurrence (BCR) using a Cox regression model, Kaplan Meier (KM) curves. We applied the bootstrapping resampling for the uncertainty evaluation and C-statistics for the randomness measurement. Results: The feature scores were significantly correlated with the androgen receptor protein expression, an androgen-signaling score, mRNA expression, and androgen receptor splice variant 7. In addition, feature scores were associated with SPINK1 overexpression, the heterozygous loss of TP53, and SPOP mutations. Additionally, the mRNA and miRNA clusters identified by the TCGA research team for PCa. These features were independent of Gleason grade and were non-random. The survival analyses revealed that a model, including three of five feature scores, achieved a c-index of 0.706 (95% CI: 0.606-0.779). The KM curve showed that these risk groups based on the Cox regression model are significantly discriminative (Log-rank P-value < 0.0001). The low-risk group (n = 177) achieved a 2-year BCR-free survival rate (BFS) of 97.4% (95% CI: 94.9 - 100.0%) and a 5-year PFS of 88.3% (95% CI: 80.6 - 96.7%). In contrast, the high-risk group (n = 194) showed a 2-year PFS of 86.3% (95% CI: 81.1 - 91.8%) and a 5-year BFS of 66.9% (95% CI: 54.6 - 0.82.1%). Conclusions: Our findings uncover the potential of feature scores from histology images as digital biomarkers in precision medicine and as an expanding utility for digital pathology.

scholarly journals Comparison of deep learning with regression analysis in creating predictive models for SARS-CoV-2 outcomes

2020 ◽  
Author(s):  
Ahmed Abdulaal ◽  
Aatish Patel ◽  
Esmita Charani ◽  
Sarah Denny ◽  
Saleh A Alqahtani ◽  
...  
Keyword(s):  
Deep Learning ◽  
Regression Model ◽  
Predictive Models ◽  
Cox Regression ◽  
Cox Model ◽  
Machine Learning Algorithms ◽  
Prognostic Models ◽  
Ann Model ◽  
Hospital Data ◽  
Cox Regression Model

Abstract Background: Accurately predicting patient outcomes in SARS-CoV-2 could aid patient management and allocation of healthcare resources. There are a variety of methods which can be used to develop prognostic models, ranging from logistic regression and survival analysis to more complex machine learning algorithms and deep learning. Despite several models having been created for SARS-CoV-2, most of these have been found to be highly susceptible to bias. We aimed to develop and compare two separate predictive models for death during admission with SARS-CoV-2. Method: Between March 1 - April 24, 2020, 398 patients were identified with laboratory confirmed SARS-CoV-2 in a London teaching hospital. Data from electronic health records were extracted and used to create two predictive models using: 1) a Cox regression model and 2) an artificial neural network (ANN). Model performance profiles were assessed by validation, discrimination, and calibration. Results: Both the Cox regression and ANN models achieved high accuracy (83.8%, 95% confidence interval (CI): 73.8 - 91.1 and 90.0%, 95% CI: 81.2 - 95.6, respectively). The area under the receiver operator curve (AUROC) for the ANN (92.6%, 95% CI: 91.1 - 94.1) was significantly greater than that of the Cox regression model (86.9%, 95% CI: 85.7 - 88.2), p=0.0136. Both models achieved acceptable calibration with Brier scores of 0.13 and 0.11 for the Cox model and ANN, respectively. Conclusion: We demonstrate an ANN which is non-inferior to a Cox regression model but with potential for further development such that it can learn as new data becomes available. Deep learning techniques are particularly suited to complex datasets with non-linear solutions, which make them appropriate for use in conditions with a paucity of prior knowledge. Accurate prognostic models for SARS-CoV-2 can provide benefits at the patient, departmental and organisational level.

scholarly journals Comparison of deep learning with regression analysis in creating predictive models for SARS-CoV-2 outcomes

2020 ◽  
Author(s):  
Ahmed Abdulaal ◽  
Aatish Patel ◽  
Esmita Charani ◽  
Sarah Denny ◽  
Saleh Alqahtani ◽  
...  
Keyword(s):  
Deep Learning ◽  
Regression Model ◽  
Predictive Models ◽  
Cox Regression ◽  
Cox Model ◽  
Machine Learning Algorithms ◽  
Prognostic Models ◽  
Ann Model ◽  
Hospital Data ◽  
Cox Regression Model

Abstract Background Accurately predicting patient outcomes in Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could aid patient management and allocation of healthcare resources. There are a variety of methods which can be used to develop prognostic models, ranging from logistic regression and survival analysis to more complex machine learning algorithms and deep learning. Despite several models having been created for SARS-CoV-2, most of these have been found to be highly susceptible to bias. We aimed to develop and compare two separate predictive models for death during admission with SARS-CoV-2.MethodBetween March 1 - April 24, 2020, 398 patients were identified with laboratory confirmed SARS-CoV-2 in a London teaching hospital. Data from electronic health records were extracted and used to create two predictive models using: 1) a Cox regression model and 2) an artificial neural network (ANN). Model performance profiles were assessed by validation, discrimination, and calibration.Results Both the Cox regression and ANN models achieved high accuracy (83.8%, 95% confidence interval (CI): 73.8 - 91.1 and 90.0%, 95% CI: 81.2 - 95.6, respectively). The area under the receiver operator curve (AUROC) for the ANN (92.6%, 95% CI: 91.1 - 94.1) was significantly greater than that of the Cox regression model (86.9%, 95% CI: 85.7 - 88.2), p=0.0136. Both models achieved acceptable calibration with Brier scores of 0.13 and 0.11 for the Cox model and ANN, respectively. ConclusionWe demonstrate an ANN which is non-inferior to a Cox regression model but with potential for further development such that it can learn as new data becomes available. Deep learning techniques are particularly suited to complex datasets with non-linear solutions, which make them appropriate for use in conditions with a paucity of prior knowledge. Accurate prognostic models for SARS-CoV-2 can provide benefits at the patient, departmental and organisational level.

scholarly journals Comparison of deep learning with regression analysis in creating predictive models for SARS-CoV-2 outcomes

2020 ◽  
Vol 20 (1) ◽  
Author(s):  
Ahmed Abdulaal ◽  
Aatish Patel ◽  
Esmita Charani ◽  
Sarah Denny ◽  
Saleh A. Alqahtani ◽  
...  
Keyword(s):  
Deep Learning ◽  
Regression Model ◽  
Predictive Models ◽  
Cox Regression ◽  
Cox Model ◽  
Machine Learning Algorithms ◽  
Prognostic Models ◽  
Ann Model ◽  
Hospital Data ◽  
Cox Regression Model

Abstract Background Accurately predicting patient outcomes in Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could aid patient management and allocation of healthcare resources. There are a variety of methods which can be used to develop prognostic models, ranging from logistic regression and survival analysis to more complex machine learning algorithms and deep learning. Despite several models having been created for SARS-CoV-2, most of these have been found to be highly susceptible to bias. We aimed to develop and compare two separate predictive models for death during admission with SARS-CoV-2. Method Between March 1 and April 24, 2020, 398 patients were identified with laboratory confirmed SARS-CoV-2 in a London teaching hospital. Data from electronic health records were extracted and used to create two predictive models using: (1) a Cox regression model and (2) an artificial neural network (ANN). Model performance profiles were assessed by validation, discrimination, and calibration. Results Both the Cox regression and ANN models achieved high accuracy (83.8%, 95% confidence interval (CI) 73.8–91.1 and 90.0%, 95% CI 81.2–95.6, respectively). The area under the receiver operator curve (AUROC) for the ANN (92.6%, 95% CI 91.1–94.1) was significantly greater than that of the Cox regression model (86.9%, 95% CI 85.7–88.2), p = 0.0136. Both models achieved acceptable calibration with Brier scores of 0.13 and 0.11 for the Cox model and ANN, respectively. Conclusion We demonstrate an ANN which is non-inferior to a Cox regression model but with potential for further development such that it can learn as new data becomes available. Deep learning techniques are particularly suited to complex datasets with non-linear solutions, which make them appropriate for use in conditions with a paucity of prior knowledge. Accurate prognostic models for SARS-CoV-2 can provide benefits at the patient, departmental and organisational level.

scholarly journals Prognostic and Predictive Value of m6A “Eraser” Related Gene Signature in Gastric Cancer

2021 ◽  
Vol 11 ◽  
Author(s):  
Xin Xu ◽  
En Zhou ◽  
Jun Zheng ◽  
Chihao Zhang ◽  
Yinghua Zou ◽  
...  
Keyword(s):  
Gastric Cancer ◽  
Prognostic Factors ◽  
Regression Model ◽  
Cox Regression ◽  
Risk Group ◽  
Critical Role ◽  
Predictive Tool ◽  
Cox Regression Model ◽  
Risk Genes ◽  
Related Risk

BackgroundN6-methyladenosine (m6A) RNA modification plays a critical role in gastric cancer (GC). However, the relationship between the m6A “eraser”, FTO, and ALKBH5, and the prognosis of GC still remains unclear. This study aimed to evaluate the effect of FTO and ALKBH5 on the prognosis of patients and their potential roles in GC.Materials and MethodsA total of 738 GC samples with clinical information obtained from two independent datasets were included and divided into training set and testing set. Differential expression analysis of the m6A “eraser” related genes was performed. The LASSO Cox regression model was constructed to analyze the m6A “eraser” related risk genes. The univariate and multivariate Cox regression model were employed to identify the independent prognostic factors. Kaplan-Meier method was used for survival analysis. A nomogram model was then carried out to predict the prognosis of GC patients. Additionally, GO and KEGG analyses were conducted to identify the potential role of the m6A “eraser” related genes in GC. The relative proportion of 22 different genotypes in immune infiltrating cells was calculated by CIBERSORT algorithm.ResultsIn total, nine m6A “eraser” related risk genes and risk scores were obtained and calculated. Patients in high-risk group demonstrated significantly worse prognosis than those in low-risk group. Age, stage, and risk score were considered as independent prognostic factors. The nomogram model constructed accurately predicted the 3-year and 5-year overall survival (OS) of patients. Furthermore, m6A “eraser” might play a functional role in GC. The expression of m6A “eraser” leads to changes in tumor immune microenvironment.ConclusionsFTO and ALKBH5 showed association with the prognosis of GC. The m6A “eraser” related genes, which is considered as a reliable prognostic and predictive tool, assists in predicting the OS in GC patients.

scholarly journals Comparison of deep learning with regression analysis in creating predictive models for SARS-CoV-2 outcomes

2020 ◽  
Author(s):  
Ahmed Abdulaal ◽  
Aatish Patel ◽  
Esmita Charani ◽  
Sarah Denny ◽  
Saleh A Alqahtani ◽  
...  
Keyword(s):  
Deep Learning ◽  
Regression Model ◽  
Predictive Models ◽  
Cox Regression ◽  
Cox Model ◽  
Machine Learning Algorithms ◽  
Prognostic Models ◽  
Ann Model ◽  
Hospital Data ◽  
Cox Regression Model

Abstract Background Accurately predicting patient outcomes in Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could aid patient management and allocation of healthcare resources. There are a variety of methods which can be used to develop prognostic models, ranging from logistic regression and survival analysis to more complex machine learning algorithms and deep learning. Despite several models having been created for SARS-CoV-2, most of these have been found to be highly susceptible to bias. We aimed to develop and compare two separate predictive models for death during admission with SARS-CoV-2.Method Between March 1 - April 24, 2020, 398 patients were identified with laboratory confirmed SARS-CoV-2 in a London teaching hospital. Data from electronic health records were extracted and used to create two predictive models using: 1) a Cox regression model and 2) an artificial neural network (ANN). Model performance profiles were assessed by validation, discrimination, and calibration.Results Both the Cox regression and ANN models achieved high accuracy (83.8%, 95% confidence interval (CI): 73.8 - 91.1 and 90.0%, 95% CI: 81.2 - 95.6, respectively). The area under the receiver operator curve (AUROC) for the ANN (92.6%, 95% CI: 91.1 - 94.1) was significantly greater than that of the Cox regression model (86.9%, 95% CI: 85.7 - 88.2), p=0.0136. Both models achieved acceptable calibration with Brier scores of 0.13 and 0.11 for the Cox model and ANN, respectively. Conclusion We demonstrate an ANN which is non-inferior to a Cox regression model but with potential for further development such that it can learn as new data becomes available. Deep learning techniques are particularly suited to complex datasets with non-linear solutions, which make them appropriate for use in conditions with a paucity of prior knowledge. Accurate prognostic models for SARS-CoV-2 can provide benefits at the patient, departmental and organisational level.

scholarly journals A Robust Seven-Gene Signature Associated With Tumor Microenvironment to Predict Survival Outcomes of Patients With Stage III–IV Lung Adenocarcinoma

2021 ◽  
Vol 12 ◽  
Author(s):  
Hao Zhao ◽  
Xuening Zhang ◽  
Lan Guo ◽  
Songhe Shi ◽  
Ciyong Lu
Keyword(s):  
High Risk ◽  
Tumor Microenvironment ◽  
Lung Adenocarcinoma ◽  
Regression Model ◽  
Cox Regression ◽  
Risk Group ◽  
External Validation ◽  
Gene Signature ◽  
Survival Outcomes ◽  
Cox Regression Model

BackgroundDue to the relatively insidious early symptoms of lung adenocarcinoma (LUAD), most LUAD patients are at an advanced stage at the time of diagnosis and lose the best chance of surgical resection. Mounting evidence suggested that the tumor microenvironment (TME) was highly correlated with tumor occurrence, progress, and prognosis. However, TME in advanced LUAD remained to be studied and reliable prognostic signatures based on TME in advanced LUAD also had not been well-established. This study aimed to understand the cell composition and function of TME and construct a gene signature associated with TME in advanced LUAD.MethodsThe immune, stromal, and ESTIMATE scores of each sample from The Cancer Genome Atlas (TCGA) database were, respectively, calculated using an ESTIMATE algorithm. The LASSO and Cox regression model were applied to select prognostic genes and to construct a gene signature associated with TME. Two independent datasets from the Gene Expression Omnibus (GEO) were used for external validation. Twenty-two subsets of tumor-infiltrating immune cells (Tiics) were analyzed using the CIBERSORT algorithm.ResultsFavorable overall survival (OS) and progression-free survival (PFS) were found in patients with high immune score (p = 0.048 and p = 0.028; respectively) and stromal score (p = 0.024 and p = 0.025; respectively). Based on the immune and stromal scores, 453 differentially expressed genes (DEGs) were identified. Using the LASSO and Cox regression model, a seven-gene signature containing AFAP1L2, CAMK1D, LOXL2, PIK3CG, PLEKHG1, RARRES2, and SPP1 was identified to construct a risk stratification model. The OS and PFS of the high-risk group were significantly worse than that of the low-risk group (p &lt; 0.001 and p &lt; 0.001; respectively). The receiver operating characteristic (ROC) curve analysis confirmed the good potency of the seven-gene signature. Similar findings were validated in two independent cohorts. In addition, the proportion of macrophages M2 and Tregs was higher in high-risk patients (p = 0.041 and p = 0.022, respectively).ConclusionOur study established and validated a seven-gene signature associated with TME, which might serve as a prognosis stratification tool to predict survival outcomes of advanced LUAD patients. In addition, macrophages M2 polarization may lead to worse prognosis in patients with advanced LUAD.

scholarly journals Construction and validation of a robust epigenetic gene-set based signature in ovarian cancer

2020 ◽  
Author(s):  
Wenbin Shen ◽  
Wei Jiang ◽  
Shuang Ye ◽  
Min Sun ◽  
Huijuan Yang ◽  
...  
Keyword(s):  
Ovarian Cancer ◽  
Regression Model ◽  
Cox Regression ◽  
Risk Group ◽  
External Validation ◽  
Epigenetic Factors ◽  
Cox Regression Model ◽  
Internal Validation ◽  
Gene Set ◽  
Validation Set

Abstract Background Epigenetic factors play a critical role in tumor development and progression. The aim of this study was to construct and validate a robust epigenetic gene-set based signature for predicting prognosis of ovarian cancer (OC). Methods Public microarray data of OC patients from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were identified and patients from TCGA dataset were randomized 3:1 into discovery and internal validation series. GSE14764 and GSE26712 from GEO database were combined as the external validation set. LASSO Cox regression model was performed in the discovery set to filter the most useful prognostic epigenetic factors. Results Based on LASSO Cox regression model, we built a 26 epigenetic factors based prognostic signature. In the discovery set, patients in high risk group showed significantly poorer overall survival than that patients in low risk group (HR: 2.11, 95% CI: 1.65–2.72, P < 0.001). The results were further validated in the internal validation set (HR: 1.69, 95% CI: 1.07–2.63, P = 0.020) and external validation set (HR 1.95, 95% CI 1.41–2.69; p < 0.001). Survival ROC at 5 year suggested that the epigenetic signature (AUC = 0.700) had better prognostic accuracy than any other clinicopathological factors in the entire cohort. In addition, survival decision curve analysis unveiled a considerable value of clinical utility of the epigenetic signature. Conclusions We successfully developed a robust epigenetic signature that can accurately predict prognosis in OC.

scholarly journals Comparison of deep learning with regression analysis in creating predictive models for SARS-CoV-2 outcomes

2020 ◽  
Author(s):  
Ahmed Abdulaal ◽  
Aatish Patel ◽  
Esmita Charani ◽  
Sarah Denny ◽  
Saleh A Alqahtani ◽  
...  
Keyword(s):  
Deep Learning ◽  
Regression Model ◽  
Predictive Models ◽  
Cox Regression ◽  
Cox Model ◽  
Machine Learning Algorithms ◽  
Prognostic Models ◽  
Ann Model ◽  
Hospital Data ◽  
Cox Regression Model

Abstract Background Accurately predicting patient outcomes in Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) could aid patient management and allocation of healthcare resources. There are a variety of methods which can be used to develop prognostic models, ranging from logistic regression and survival analysis to more complex machine learning algorithms and deep learning. Despite several models having been created for SARS-CoV-2, most of these have been found to be highly susceptible to bias. We aimed to develop and compare two separate predictive models for death during admission with SARS-CoV-2.MethodBetween March 1 - April 24, 2020, 398 patients were identified with laboratory confirmed SARS-CoV-2 in a London teaching hospital. Data from electronic health records were extracted and used to create two predictive models using: 1) a Cox regression model and 2) an artificial neural network (ANN). Model performance profiles were assessed by validation, discrimination, and calibration.Results Both the Cox regression and ANN models achieved high accuracy (83.8%, 95% confidence interval (CI): 73.8 - 91.1 and 90.0%, 95% CI: 81.2 - 95.6, respectively). The area under the receiver operator curve (AUROC) for the ANN (92.6%, 95% CI: 91.1 - 94.1) was significantly greater than that of the Cox regression model (86.9%, 95% CI: 85.7 - 88.2), p=0.0136. Both models achieved acceptable calibration with Brier scores of 0.13 and 0.11 for the Cox model and ANN, respectively. ConclusionWe demonstrate an ANN which is non-inferior to a Cox regression model but with potential for further development such that it can learn as new data becomes available. Deep learning techniques are particularly suited to complex datasets with non-linear solutions, which make them appropriate for use in conditions with a paucity of prior knowledge. Accurate prognostic models for SARS-CoV-2 can provide benefits at the patient, departmental and organisational level.

scholarly journals 813. The Reduction of the acquisition rate of carbapenem resistant Acinetobacter baumannii after room privatization in the intensive care unit

2020 ◽  
Vol 7 (Supplement_1) ◽  
pp. S448-S449
Author(s):  
Jongtak Jung ◽  
Pyoeng Gyun Choe ◽  
Chang Kyung Kang ◽  
Kyung Ho Song ◽  
Wan Beom Park ◽  
...  
Keyword(s):  
Regression Model ◽  
Acinetobacter Baumannii ◽  
Cox Regression ◽  
Univariate Analysis ◽  
Intervention Group ◽  
Feeding Tube ◽  
Control Group ◽  
Cox Regression Model ◽  
Carbapenem Resistant ◽  
Hospital Acquired

Abstract Background Acinetobacter baumannii is one of the major pathogens of hospital-acquired infection recently and hospital outbreaks have been reported worldwide. On September 2017, New intensive care unit(ICU) with only single rooms, remodeling from old ICU with multibed bay rooms, was opened in an acute-care tertiary hospital in Seoul, Korea. We investigated the effect of room privatization in the ICU on the acquisition of carbapenem-resistant Acinetobacter baumannii(CRAB). Methods We retrospectively reviewed medical records of patients who admitted to the medical ICU in a tertiary care university-affiliated 1,800-bed hospital from 1 January 2015 to 1 January 2019. Patients admitted to the medical ICU before the remodeling of the ICU were designated as the control group, and those who admitted to the medical ICU after the remodeling were designated as the intervention group. Then we compared the acquisition rate of CRAB between the control and intervention groups. Patients colonized with CRAB or patients with CRAB identified in screening tests were excluded from the study population. The multivariable Cox regression model was performed using variables with p-values of less than 0.1 in the univariate analysis. Results A total of 1,105 cases admitted to the ICU during the study period were analyzed. CRAB was isolated from 110 cases in the control group(n=687), and 16 cases in the intervention group(n=418). In univariate analysis, room privatization, prior exposure to antibiotics (carbapenem, vancomycin, fluoroquinolone), mechanical ventilation, central venous catheter, tracheostomy, the presence of feeding tube(Levin tube or percutaneous gastrostomy) and the length of ICU stay were significant risk factors for the acquisition of CRAB (p&lt; 0.05). In the multivariable Cox regression model, the presence of feeding tube(Hazard ratio(HR) 4.815, 95% Confidence interval(CI) 1.94-11.96, p=0.001) and room privatization(HR 0.024, 95% CI 0.127-0.396, p=0.000) were independent risk factors. Table 1. Univariate analysis of Carbapenem-resistant Acinetobacter baumannii Table 2. Multivariable Cox regression model of the acquisition of Carbapenem-resistant Acinetobacter baumannii Conclusion In the present study, room privatization of the ICU was correlated with the reduction of CRAB acquisition independently. Remodeling of the ICU to the single room would be an efficient strategy for preventing the spreading of multidrug-resistant organisms and hospital-acquired infection. Disclosures All Authors: No reported disclosures

Sign in / Sign up

Export Citation Format

Share Document