Heterogeneous ensembles for predicting survival of metastatic, castrate-resistant prostate cancer patients

CT radiomics model combined with clinical and radiographic features for discriminating peripheral small cell lung cancer from peripheral lung adenocarcinoma

Purpose

Exploring a non-invasive method to accurately differentiate peripheral small cell lung cancer (PSCLC) and peripheral lung adenocarcinoma (PADC) could improve clinical decision-making and prognosis.

Methods

This retrospective study reviewed the clinicopathological and imaging data of lung cancer patients between October 2017 and March 2022. A total of 240 patients were enrolled in this study, including 80 cases diagnosed with PSCLC and 160 with PADC. All patients were randomized in a seven-to-three ratio into the training and validation datasets (170 vs. 70, respectively). The least absolute shrinkage and selection operator regression was employed to generate radiomics features and univariate analysis, followed by multivariate logistic regression to select significant clinical and radiographic factors to generate four models: clinical, radiomics, clinical-radiographic, and clinical-radiographic-radiomics (comprehensive). The Delong test was to compare areas under the receiver operating characteristic curves (AUCs) in the models.

Results

Five clinical-radiographic features and twenty-three selected radiomics features differed significantly in the identification of PSCLC and PADC. The clinical, radiomics, clinical-radiographic and comprehensive models demonstrated AUCs of 0.8960, 0.8356, 0.9396, and 0.9671 in the validation set, with the comprehensive model having better discernment than the clinical model (P=0.036), the radiomics model (P=0.006) and the clinical-radiographic model (P=0.049).

Conclusions

The proposed model combining clinical data, radiographic characteristics and radiomics features could accurately distinguish PSCLC from PADC, thus providing a potential non-invasive method to help clinicians improve treatment decisions.

Machine learning based survival prediction in Glioma using large-scale registry data

Gliomas are the most common central nervous system tumors exhibiting poor clinical outcomes. The ability to estimate prognosis is crucial for both patients and providers in order to select the most appropriate treatment. Machine learning (ML) allows for sophisticated approaches to survival prediction using real world clinical parameters needed to achieve superior predictive accuracy. We employed Cox Proportional hazards (CPH) model, Support Vector Machine (SVM) model, Random Forest (RF) model in a large glioma dataset (3462 patients, diagnosed 2000-2018) to explore the most optimal approach to survival prediction. Features employed were age, sex, surgical resection status, tumor histology and tumor site, administration of radiation therapy (RT) and chemotherapy status. Concordance index (c-index) was employed to assess the accuracy of survival time prediction. All three models performed well with prediction accuracy (CI 0.767, 0.771, 0.57 for CPH, SVM, RF models respectively) with the best performance achieved when incorporating RT and chemotherapy administration status which emerged as key predictive features. Within the subset of glioblastoma patients, similar prediction accuracy was achieved. These findings should prompt stricter clinician oversight over registry data accuracy through quality assurance as we move towards meaningful predictive ability using ML approaches in glioma.

Progression-Free Survival Prediction in Small Cell Lung Cancer Based on Radiomics Analysis of Contrast-Enhanced CT

Purposes and Objectives

The aim of this study was to predict the progression-free survival (PFS) in patients with small cell lung cancer (SCLC) by radiomic signature from the contrast-enhanced computed tomography (CT).

Methods

A total of 186 cases with pathological confirmed small cell lung cancer were retrospectively assembled. First, 1,218 radiomic features were automatically extracted from tumor region of interests (ROIs) on the lung window and mediastinal window, respectively. Then, the prognostic and robust features were selected by machine learning methods, such as (1) univariate analysis based on a Cox proportional hazard (CPH) model, (2) redundancy removing using the variance inflation factor (VIF), and (3) multivariate importance analysis based on random survival forests (RSF). Finally, PFS predictive models were established based on RSF, and their performances were evaluated using the concordance index (C-index) and the cumulative/dynamic area under the curve (C/D AUC).

Results

In total, 11 radiomic features (6 for mediastinal window and 5 for lung window) were finally selected, and the predictive model constructed from them achieved a C-index of 0.7531 and a mean C/D AUC of 0.8487 on the independent test set, better than the predictions by single clinical features (C-index = 0.6026, mean C/D AUC = 0.6312), and single radiomic features computed in lung window (C-index = 0.6951, mean C/D AUC = 0.7836) or mediastinal window (C-index = 0.7192, mean C/D AUC = 0.7964).

Conclusion

The radiomic features computed from tumor ROIs on both lung window and mediastinal window can predict the PFS for patients with SCLC by a high accuracy, which could be used as a useful tool to support the personalized clinical decision for the diagnosis and patient management of patients with SCLC.

Can machine learning bring cardiovascular risk assessment to the next level? A methodological study using FOURIER trial data

Aims

Through this proof of concept, we studied the potential added value of machine learning (ML) methods in building cardiovascular risk scores from structured data and the conditions under which they outperform linear statistical models.

Methods and results

Relying on extensive cardiovascular clinical data from FOURIER, a randomized clinical trial to test for evolocumab efficacy, we compared linear models, neural networks, random forest, and gradient boosting machines for predicting the risk of major adverse cardiovascular events. To study the relative strengths of each method, we extended the comparison to restricted subsets of the full FOURIER dataset, limiting either the number of available patients or the number of their characteristics. When using all the 428 covariates available in the dataset, ML methods significantly (c-index 0.67, P-value 2e-5) outperformed linear models built from the same variables (c-index 0.62), as well as a reference cardiovascular risk score based on only 10 variables (c-index 0.60). We showed that gradient boosting-the best performing model in our setting-requires fewer patients and significantly outperforms linear models when using large numbers of variables. On the other hand, we illustrate how linear models suffer from being trained on too many variables, thus requiring a more careful prior selection. These ML methods proved to consistently improve risk assessment, to be interpretable despite their complexity and to help identify the minimal set of covariates necessary to achieve top performance.

Conclusion

In the field of secondary cardiovascular events prevention, given the increased availability of extensive electronic health records, ML methods could open the door to more powerful tools for patient risk stratification and treatment allocation strategies.

Radiomics model of 18F-FDG PET/CT imaging for predicting disease-free survival of early-stage uterine cervical squamous cancer

BACKGROUND

To explore an effective predictive model based on PET/CT radiomics for the prognosis of early-stage uterine cervical squamous cancer.

METHODS

Preoperative PET/CT data were collected from 201 uterine cervical squamous cancer patients with stage IB-IIA disease (FIGO 2009) who underwent radical surgery between 2010 and 2015. The tumor regions were manually segmented, and 1318 radiomic features were extracted. First, model-based univariate analysis was performed to exclude features with small correlations. Then, the redundant features were further removed by feature collinearity. Finally, the random survival forest (RSF) was used to assess feature importance for multivariate analysis. The prognostic models were established based on RSF, and their predictive performances were measured by the C-index and the time-dependent cumulative/dynamics AUC (C/D AUC).

RESULTS

In total, 6 radiomic features (5 for CT and 1 for PET) and 6 clinicopathologic features were selected. The radiomic, clinicopathologic and combination prognostic models yielded C-indexes of 0.9338, 0.9019 and 0.9527, and the mean values of the C/D AUC (mC/D AUC) were 0.9146, 0.8645 and 0.9199, respectively.

CONCLUSIONS

PET/CT radiomics could achieve approval power in predicting DFS in early-stage uterine cervical squamous cancer.

Identifying Prognostic Markers From Clinical, Radiomics, and Deep Learning Imaging Features for Gastric Cancer Survival Prediction

BACKGROUND

Gastric cancer is one of the leading causes of cancer death in the world. Improving gastric cancer survival prediction can enhance patient prognostication and treatment planning.

METHODS

In this study, we performed gastric cancer survival prediction using machine learning and multi-modal data of 1061 patients, including 743 for model learning and 318 independent patients for evaluation. A Cox proportional-hazard model was trained to integrate clinical variables and CT imaging features (extracted by radiomics and deep learning) for overall and progression-free survival prediction. We further analyzed the prediction effects of clinical, radiomics, and deep learning features. Concordance index (c-index) was used as the model performance metric, and the predictive effects of multi-modal features were measured by hazard ratios (HRs) at pre- and post-operative settings.

RESULTS

Among 318 patients in the independent testing group, the hazard predicted by Cox from multi-modal features is associated with their survival. The highest c-index was 0.783 (95% CI, 0.782-0.783) and 0.770 (95% CI, 0.769-0.771) for overall and progression-free survival prediction, respectively. The post-operative variables are significantly (p<0.001) more predictive than the pre-operative variables. Pathological tumor stage (HR=1.336 [overall survival]/1.768 [progression-free survival], p<0.005), pathological lymph node stage (HR=1.665/1.433, p<0.005), carcinoembryonic antigen (CEA) (HR=1.632/1.522, p=0.02), chemotherapy treatment (HR=0.254/0.287, p<0.005), radiomics signature [HR=1.540/1.310, p<0.005], and deep learning signature [HR=1.950/1.420, p<0.005]) are significant survival predictors.

CONCLUSION

Our study showed that CT radiomics and deep learning imaging features are significant pre-operative predictors, providing additional prognostic information to the pathological staging markers. Lower CEA levels and chemotherapy treatments also increase survival chances. These findings can enhance gastric cancer patient prognostication and inform treatment planning.

Machine-Learning Approaches in COVID-19 Survival Analysis and Discharge-Time Likelihood Prediction Using Clinical Data

As a highly contagious respiratory disease, COVID-19 has yielded high mortality rates since its emergence in December 2019. As the number of COVID-19 cases soars in epicenters, health officials are warning about the possibility of the designated treatment centers being overwhelmed by coronavirus patients. In this study, several computational techniques are implemented to analyze the survival characteristics of 1,182 patients. The computational results agree with the outcome reported in early clinical reports released for a group of patients from China that confirmed a higher mortality rate in men compared with women and in older age groups. The discharge-time prediction of COVID-19 patients was also evaluated using different machine-learning and statistical analysis methods. The results indicate that the Gradient Boosting survival model outperforms other models for patient survival prediction in this study. This research study is aimed to help health officials make more educated decisions during the outbreak.