scholarly journals A prospectively validated machine learning model for the prediction of survival and tumor subtype in pancreatic ductal adenocarcinoma

2019 ◽  
Author(s):  
Georgios Kaissis ◽  
Sebastian Ziegelmayer ◽  
Fabian Lohöfer ◽  
Hana Algül ◽  
Matthias Eiber ◽  
...  
Keyword(s):  
Machine Learning ◽  
Overall Survival ◽  
Median Overall Survival ◽  
Validation Cohort ◽  
Learning Algorithm ◽  
External Validation ◽  
Ductal Adenocarcinoma ◽  
Machine Learning Algorithm ◽  
Training Cohort ◽  
Histopathological Subtype

AbstractPurposeTo develop a supervised machine learning algorithm capable of predicting above vs. below-median overall survival from medical imaging-derived radiomic features in a cohort of patients with pancreatic ductal adenocarcinoma (PDAC).Materials and Methods102 patients with histopathologically proven PDAC were retrospectively assessed as the training cohort and 30 prospectively enrolled patients served as the external validation cohort. Tumors were segmented in pre-operative diffusion weighted-(DW)-MRI derived ADC maps and radiomic features were extracted. A Random Forest machine learning algorithm was fit to the training cohort and tested in the external validation cohort. The histopathological subtype of the tumor samples was assessed by immunohistochemistry in 21/30 patients of the external validation cohort. Individual radiomic feature importance was evaluated.ResultsThe machine learning algorithm achieved a sensitivity of 87% and a specificity of 80% (ROC-AUC 90%) for the prediction of above- vs. below-median survival on the unseen data of the external validation cohort. Heterogeneity-related features were highly ranked by the model. Of the 21 patients for whom the histopathological subtype was determined, 8/9 patients predicted by the model to experience below-median overall survival exhibited the quasi-mesenchymal subtype, while 11/12 patients predicted to experience above-median survival exhibited a non-quasi-mesenchymal subtype (Fisher’s exact test P<0.001).ConclusionThe application of machine-learning to the radiomic analysis of DW-MRI-derived ADC maps allowed the prediction of overall survival with high diagnostic accuracy in a prospectively collected cohort. The high overlap of clinically relevant histopathological subtypes with model predictions underlines the potential of quantitative imaging workflows in pre-operative subtyping and risk assessment in PDAC.

Abstract WP406: Training and Validation of Deepmedic Machine Learning Tool for Automated Hematoma Segmentation and Volume Analysis on CT Using Multicenter Data

Stroke ◽  
2020 ◽  
Vol 51 (Suppl_1) ◽  
Author(s):  
Sean Nurmsoo ◽  
Alessandro Guida ◽  
Alex Wong ◽  
Richard I Aviv ◽  
Andrew Demchuk ◽  
...  
Keyword(s):  
Machine Learning ◽  
Validation Cohort ◽  
Learning Algorithm ◽  
Intraclass Correlation ◽  
Ct Scans ◽  
Dice Similarity Coefficient ◽  
Machine Learning Algorithm ◽  
Manual Segmentation ◽  
Training Cohort ◽  
Volume Calculation

Introduction: We sought to train and validate an automated machine learning algorithm for ICH segmentation and volume calculation using multicenter data. Methods: An open-source 3D deep machine learning algorithm “DeepMedic” was trained using manually segmented ICH from 208 CT scans (129 patients) from the multicenter PREDICT study. The algorithm was then validated with 125 manually segmented CT scans (48 patients) from the SPOTLIGHT study. Manual segmentation was performed with Quantomo semi-automated software. ABC/2 was measured for all studies by two neuroradiologists. Accuracy of DeepMedic segmentation was assessed using the Dice similarity coefficient. Analysis was stratified by presence of IVH. Intraclass correlation (ICC) with 95% confidence intervals (CI) assessed agreement between manual vs. DeepMedic segmentation volume; and manual segmentation and ABC/2 volume. Bland-Altman charts were analyzed for ABC/2 and DeepMedic vs. manual segmentation volumes. Results: DeepMedic demonstrated high segmentation accuracy in the training cohort (median Dice 0.96; IQR 0.95 - 0.97) and in the validation cohort (median Dice 0.91; IQR 0.86 - 0.94). Dice coefficients were not significantly different between patients with IVH in the training cohort; however was significantly worse in the validation cohort in patients with IVH (Wilcoxon p<0.001). Agreement was significantly better between DeepMedic and manual segmentation (PREDICT: ICC 0.99 [95%CI 0.99 -1.00]; SPOTLIGHT: ICC 0.98 [95%CI 0.97 - 0.99]) than between ABC/2 and manual segmentation (PREDICT: ICC 0.92 [95%CI 0.89 - 0.95]; SPOTLIGHT: ICC 0.95 [95%CI 0.93-0.97]). Improved accuracy of DeepMedic was demonstrated in Bland-Altman charts (Fig 1). Conclusion: ICH machine learning segmentation with DeepMedic is feasible and accurate; and demonstrates greater agreement with manual segmentation compared to ABC/2 volumes. Accuracy of the machine learning algorithm however is limited in patients with IVH.

scholarly journals Interpretable Machine Learning Algorithm Reveals Novel Gut Microbiome Features in Predicting Type 2 Diabetes

2020 ◽  
Vol 4 (Supplement_2) ◽  
pp. 1559-1559
Author(s):  
Wanglong Gou ◽  
Chu-Wen Ling ◽  
Yan He ◽  
Zengliang Jiang ◽  
Yuanqing Fu ◽  
...  
Keyword(s):  
Machine Learning ◽  
Type 2 Diabetes ◽  
Gut Microbiome ◽  
Large Scale ◽  
Validation Cohort ◽  
Learning Algorithm ◽  
External Validation ◽  
Positive Association ◽  
Interpretable Machine Learning

Abstract Objectives The gut microbiome-type 2 diabetes (T2D) relationship among human cohorts have been controversial. We hypothesized that this limitation could be addressed by integrating the cutting-edge interpretable machine learning framework and large-scale human cohort studies. Methods 3 independent cohorts with &gt;9000 participants were included in this study. We proposed a new machine learning-based analytic framework — using LightGBM to infer the relationship between incorporated features and T2D, and SHapley Additive explanation(SHAP) to identified microbiome features associated with the risk of T2D. We then generated a microbiome risk score (MRS) integrating the threshold and direction of the identified microbiome features to predict T2D risk. Results We finally identified 15 microbiome features (two of them are indicators of microbial diversity, others are taxa-related features) associated with the risk of T2D. The identified T2D-related gut microbiome features showed superior T2D prediction accuracy compared to host genetics or traditional risk factors. Furthermore, we found that the MRS (per unit change in MRS) consistently showed positive association with T2D risk in the discovery cohort (RR 1.28, 95%CI 1.23-1.33), external validation cohort 1 (RR 1.23, 95%CI 1.13-1.34) and external validation cohort 2 (GGMP, RR 1.12, 95%CI 1.06-1.18). The MRS could also predict future glucose increment. We subsequently identified dietary and lifestyle factors which could prospectively modulate the microbiome features, and found that body fat distribution may be the key factor modulating the gut microbiome-T2D relationship. Conclusions Taken together, we proposed a new analytical framework for the investigation of microbiome-disease relationship. The identified microbiome features may serve as potential drug targets for T2D in future. Funding Sources This study was funded by National Natural Science Foundation of China (81903316, 81773416), Westlake University (101396021801) and the 5010 Program for Clinical Researches (2007032) of the Sun Yat-sen University (Guangzhou, China).

scholarly journals A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy

2019 ◽  
Author(s):  
Georgios Kaissis ◽  
Sebastian Ziegelmayer ◽  
Fabian Lohöfer ◽  
Katja Steiger ◽  
Hana Algül ◽  
...  
Keyword(s):  
Machine Learning ◽  
Overall Survival ◽  
Pancreatic Ductal Adenocarcinoma ◽  
Median Overall Survival ◽  
Molecular Subtypes ◽  
Supervised Machine Learning ◽  
Ductal Adenocarcinoma ◽  
Chemotherapy Response ◽  
Response To Chemotherapy ◽  
Sensitivity Specificity

AbstractPurposeDevelopment of a supervised machine-learning model capable of predicting clinically relevant molecular subtypes of pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features.MethodsThe retrospective observational study assessed 55 surgical PDAC patients. Molecular subtypes were defined by immunohistochemical staining of KRT81. Tumors were manually segmented and 1606 radiomic features were extracted withPyRadiomics. A gradient-boosted-tree algorithm (XGBoost) was trained on 70% of the patients (N=28) and tested on 30% (N=17) to predict KRT81+ vs. KRT81-tumor subtypes. The average sensitivity, specificity and ROC-AUC value were calculated. Chemotherapy response was assessed stratified by subtype. Radiomic feature importance was ranked.ResultsThe mean±STDEV sensitivity, specificity and ROC-AUC were 0.90±0.07, 0.92±0.11, and 0.93±0.07, respectively. Patients with a KRT81+ subtype experienced significantly diminished median overall survival compared to KRT81-patients (7.0 vs. 22.6 months, HR 1.44, log-rank-test P=<0.001) and a significantly improved response to gemcitabine-based chemotherapy over FOLFIRINOX (10.14 vs. 3.8 months median overall survival, HR 0.85, P=0.037) compared to KRT81-patients, who responded significantly better to FOLFIRINOX over gemcitabine-based treatment (30.8 vs. 13.4 months median overall survival, HR 0.88, P=0.027).ConclusionsThe machine-learning based analysis of radiomic features enables the prediction of subtypes of PDAC, which are highly relevant for overall patient survival and response to chemotherapy.

scholarly journals Cathepsin D Expression and Gemcitabine Resistance in Pancreatic Cancer

2019 ◽  
Vol 4 (1) ◽  
Author(s):  
Ujjwal M Mahajan ◽  
Elisabetta Goni ◽  
Enno Langhoff ◽  
Qi Li ◽  
Eithne Costello ◽  
...  
Keyword(s):  
Pancreatic Cancer ◽  
Overall Survival ◽  
Cathepsin D ◽  
Median Overall Survival ◽  
Validation Cohort ◽  
Multivariable Analysis ◽  
Predictive Marker ◽  
Acid Sphingomyelinase ◽  
Ductal Adenocarcinoma ◽  
Gemcitabine Resistance

Abstract Background Cathepsin-D (CatD), owing to its dual role as a proteolytic enzyme and as a ligand, has been implicated in cancer progression. The role of CatD in pancreatic ductal adenocarcinoma is unknown. Methods CatD expression quantified by immunohistochemistry of tumor-tissue microarrays of 403 resected pancreatic cancer patients from the ESPAC-Tplus trial, a translational study within the ESPAC (European Study Group for Pancreatic Cancer) trials, was dichotomously distributed to low and high H scores (cut off 22.35) for survival and multivariable analysis. The validation cohort (n = 69) was recruited based on the hazard ratio of CatD from ESPAC-Tplus. 5-fluorouracil-, and gemcitabine-resistant pancreatic cancer cell lines were employed for mechanistic experiments. All statistical tests were two-sided. Results Median overall survival was 23.75 months and median overall survival for patients with high CatD expression was 21.09 (95% confidence interval [CI] = 17.31 to 24.80) months vs 27.20 (95% CI = 23.75 to 31.90) months for low CatD expression (χ2LR, 1DF = 4.00; P = .04). Multivariable analysis revealed CatD expression as a predictive marker in gemcitabine-treated (z stat = 2.33; P = .02) but not in 5-fluorouracil-treated (z stat = 0.21; P = .82) patients. An independent validation cohort confirmed CatD as a negative predictive marker for survival (χ2LR, 1DF = 6.80; P = .009) and as an independent predictive marker in gemcitabine-treated patients with a hazard ratio of 3.38 (95% CI = 1.36 to 8.38, P = .008). Overexpression of CatD was associated with a concomitant suppression of the acid sphingomyelinase, and silencing of CatD resulted in upregulation of acid sphingomyelinase with rescue of gemcitabine resistance. Conclusions Adjuvant gemcitabine is less effective in pancreatic ductal adenocarcinoma with high CatD expression, and thus CatD could serve as a marker for biomarker-driven therapy.

scholarly journals Decision Tree With Only Two Musculoskeletal Sites to Diagnose Polymyalgia Rheumatica Using [18F]FDG PET-CT

2021 ◽  
Vol 8 ◽  
Author(s):  
Anthime Flaus ◽  
Julie Amat ◽  
Nathalie Prevot ◽  
Louis Olagne ◽  
Lucie Descamps ◽  
...  
Keyword(s):  
Machine Learning ◽  
Decision Tree ◽  
Polymyalgia Rheumatica ◽  
Fdg Pet ◽  
Validation Cohort ◽  
Learning Algorithm ◽  
Final Diagnosis ◽  
Machine Learning Algorithm ◽  
Pet Ct ◽  
Fdg Pet Ct

Introduction: The aim of this study was to find the best ordered combination of two FDG positive musculoskeletal sites with a machine learning algorithm to diagnose polymyalgia rheumatica (PMR) vs. other rheumatisms in a cohort of patients with inflammatory rheumatisms.Methods: This retrospective study included 140 patients who underwent [18F]FDG PET-CT and whose final diagnosis was inflammatory rheumatism. The cohort was randomized, stratified on the final diagnosis into a training and a validation cohort. FDG uptake of 17 musculoskeletal sites was evaluated visually and set positive if uptake was at least equal to that of the liver. A decision tree classifier was trained and validated to find the best combination of two positives sites to diagnose PMR. Diagnosis performances were measured first, for each musculoskeletal site, secondly for combination of two positive sites and thirdly using the decision tree created with machine learning.Results: 55 patients with PMR and 85 patients with other inflammatory rheumatisms were included. Musculoskeletal sites, used either individually or in combination of two, were highly imbalanced to diagnose PMR with a high specificity and a low sensitivity. The machine learning algorithm identified an optimal ordered combination of two sites to diagnose PMR. This required a positive interspinous bursa or, if negative, a positive trochanteric bursa. Following the decision tree, sensitivity and specificity to diagnose PMR were respectively 73.2 and 87.5% in the training cohort and 78.6 and 80.1% in the validation cohort.Conclusion: Ordered combination of two visually positive sites leads to PMR diagnosis with an accurate sensitivity and specificity vs. other rheumatisms in a large cohort of patients with inflammatory rheumatisms.

scholarly journals Pancreatic Ductal Adenocarcinoma at CT: A Combined Nomogram Model to Preoperatively Predict Cancer Stage and Survival Outcome

2021 ◽  
Vol 11 ◽  
Author(s):  
Chunyuan Cen ◽  
Liying Liu ◽  
Xin Li ◽  
Ailan Wu ◽  
Huan Liu ◽  
...  
Keyword(s):  
Overall Survival ◽  
Pancreatic Ductal Adenocarcinoma ◽  
Clinical Characteristics ◽  
Validation Cohort ◽  
Stage I ◽  
Ductal Adenocarcinoma ◽  
Training Cohort ◽  
Calibration Curves ◽  
Kaplan Meier ◽  
Radiomics Signature

ObjectivesTo construct a nomogram model that combines clinical characteristics and radiomics signatures to preoperatively discriminate pancreatic ductal adenocarcinoma (PDAC) in stage I-II and III-IV and predict overall survival.MethodsA total of 135 patients with histopathologically confirmed PDAC who underwent contrast-enhanced CT were included. A total of 384 radiomics features were extracted from arterial phase (AP) or portal venous phase (PVP) images. Four steps were used for feature selection, and multivariable logistic regression analysis were used to build radiomics signatures and combined nomogram model. Performance of the proposed model was assessed by using receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA). Kaplan-Meier analysis was applied to analyze overall survival in the stage I-II and III-IV PDAC groups.ResultsThe AP+PVP radiomics signature showed the best performance among the three radiomics signatures [training cohort: area under the curve (AUC) = 0.919; validation cohort: AUC = 0.831]. The combined nomogram model integrating AP+PVP radiomics signature with clinical characteristics (tumor location, carcinoembryonic antigen level, and tumor maximum diameter) demonstrated the best discrimination performance (training cohort: AUC = 0.940; validation cohort: AUC = 0.912). Calibration curves and DCA verified the clinical usefulness of the combined nomogram model. Kaplan-Meier analysis showed that overall survival of patients in the predicted stage I-II PDAC group was longer than patients in stage III-IV PDAC group (p&lt;0.0001).ConclusionsWe propose a combined model with excellent performance for the preoperative, individualized, noninvasive discrimination of stage I-II and III-IV PDAC and prediction of overall survival.

scholarly journals The use of a machine-learning algorithm that predicts hypotension during surgery in combination with personalized treatment guidance: study protocol for a randomized clinical trial

Trials ◽  
2019 ◽  
Vol 20 (1) ◽  
Author(s):  
M. Wijnberge ◽  
J. Schenk ◽  
L. E. Terwindt ◽  
M. P. Mulder ◽  
M. W. Hollmann ◽  
...  
Keyword(s):  
Machine Learning ◽  
Learning Algorithm ◽  
Controlled Trial ◽  
External Validation ◽  
Treatment Protocol ◽  
Weighted Average ◽  
Current Treatment ◽  
Personalized Treatment ◽  
Arterial Line ◽  
Machine Learning Algorithm

Abstract Background Intraoperative hypotension is associated with increased morbidity and mortality. Current treatment is mostly reactive. The Hypotension Prediction Index (HPI) algorithm is able to predict hypotension minutes before the blood pressure actually decreases. Internal and external validation of this algorithm has shown good sensitivity and specificity. We hypothesize that the use of this algorithm in combination with a personalized treatment protocol will reduce the time weighted average (TWA) in hypotension during surgery spent in hypotension intraoperatively. Methods/design We aim to include 100 adult patients undergoing non-cardiac surgery with an anticipated duration of more than 2 h, necessitating the use of an arterial line, and an intraoperatively targeted mean arterial pressure (MAP) of > 65 mmHg. This study is divided into two parts; in phase A baseline TWA data from 40 patients will be collected prospectively. A device (HemoSphere) with HPI software will be connected but fully covered. Phase B is designed as a single-center, randomized controlled trial were 60 patients will be randomized with computer-generated blocks of four, six or eight, with an allocation ratio of 1:1. In the intervention arm the HemoSphere with HPI will be used to guide treatment; in the control arm the HemoSphere with HPI software will be connected but fully covered. The primary outcome is the TWA in hypotension during surgery. Discussion The aim of this trial is to explore whether the use of a machine-learning algorithm intraoperatively can result in less hypotension. To test this, the treating anesthesiologist will need to change treatment behavior from reactive to proactive. Trial registration This trial has been registered with the NIH, U.S. National Library of Medicine at ClinicalTrials.gov, ID: NCT03376347. The trial was submitted on 4 November 2017 and accepted for registration on 18 December 2017.

scholarly journals Using a machine learning algorithm to predict acute graft-versus-host disease following allogeneic transplantation

2019 ◽  
Vol 3 (22) ◽  
pp. 3626-3634 ◽  
Author(s):  
Yasuyuki Arai ◽  
Tadakazu Kondo ◽  
Kyoko Fuse ◽  
Yasuhiko Shibasaki ◽  
Masayoshi Masuko ◽  
...  
Keyword(s):  
Machine Learning ◽  
Overall Survival ◽  
Risk Stratification ◽  
Learning Algorithm ◽  
Allogeneic Transplantation ◽  
Machine Learning Algorithms ◽  
Machine Learning Algorithm ◽  
Graft Versus Host ◽  
Key Points ◽  
Acute Graft

Key Points The machine learning algorithms produced clinically reasonable and robust risk stratification scores for aGVHD. Predicting scores for aGVHD also demonstrated the link between risk of development of aGVHD and overall survival after HSCT.

Sign in / Sign up

Export Citation Format

Share Document