Year-end Sale % OFF 
Abstract
Radiomics applies machine learning algorithms to quantitative imaging data to characterise the tumour phenotype and predict clinical outcome. For the development of radiomics risk models, a variety of different algorithms is available and it is not clear which one gives optimal results. Therefore, we assessed the performance of 11 machine learning algorithms combined with 12 feature selection methods by the concordance index (C-Index), to predict loco-regional tumour control (LRC) and overall survival for patients with head and neck squamous cell carcinoma. The considered algorithms are able to deal with continuous time-to-event survival data. Feature selection and model building were performed on a multicentre cohort (213 patients) and validated using an independent cohort (80 patients). We found several combinations of machine learning algorithms and feature selection methods which achieve similar results, e.g., MSR-RF: C-Index = 0.71 and BT-COX: C-Index = 0.70 in combination with Spearman feature selection. Using the best performing models, patients were stratified into groups of low and high risk of recurrence. Significant differences in LRC were obtained between both groups on the validation cohort. Based on the presented analysis, we identified a subset of algorithms which should be considered in future radiomics studies to develop stable and clinically relevant predictive models for time-to-event endpoints.
Highlights
The prognostic performance of different feature selection methods combined with machine learning algorithms was evaluated for the clinical endpoint loco-regional tumour control (LRC)
The best single performances were obtained by the MSR-RF (C-Index: 0.71, 95% confidence interval [0.62–0.83]), the boosting trees (BT)-CIndex (C-Index: 0.71, [0.62–0.82]), the BT-Weibull (C-Index: 0.70, [0.60–0.82]) and the BT-COX (C-Index: 0.70, [0.59–0.81]) algorithms, all in combination with the Spearman feature selection method
The highest single prediction performances were obtained by the boosting gradient linear models (BGLM)-CIndex (C-Index: 0.64, [0.53–0.71]), the BGLM-Weibull (C-Index: 0.64, [0.52–0.70]) and the BGLM-COX (C-Index: 0.64, [0.51–0.68]), all in combination with the random feature selection
Summary
Radiomics aims to predict patient specific outcomes based on high-throughput analysis and mining of advanced imaging biomarkers by machine learning algorithms. It has shown promising results in several studies on lung, head and neck, breast as well as brain tumours[2,3,4,5,6,7]. A systemic evaluation to identify a set of suitable feature selection methods and learning algorithms is a critical step to develop clinically relevant radiomics risk models. We assessed both, predictive performance of the models and patient risk stratification, for each combination of feature selection and learning algorithm on the validation cohort. The evaluations above led to the identification of a subset of useful feature selection and learning algorithms for time-to-event survival data
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
