Predicting clinical outcomes in COVID-19 using radiomics on chest radiographs

Abstract

Objectives: For optimal utilization of healthcare resources, there is a critical need for early identification of COVID-19 patients at risk of poor prognosis as defined by the need for intensive unit care and mechanical ventilation. We tested the feasibility of chest X-ray (CXR)-based radiomics metrics to develop machine-learning algorithms for predicting patients with poor outcomes.

Methods: In this Institutional Review Board (IRB) approved, Health Insurance Portability and Accountability Act (HIPAA) compliant, retrospective study, we evaluated CXRs performed around the time of admission from 167 COVID-19 patients. Of the 167 patients, 68 (40.72%) required intensive care during their stay, 45 (26.95%) required intubation, and 25 (14.97%) died. Lung opacities were manually segmented using ITK-SNAP (open-source software). CaPTk (open-source software) was used to perform 2D radiomics analysis.

Results: Of all the algorithms considered, the AdaBoost classifier performed the best with AUC = 0.72 to predict the need for intubation, AUC = 0.71 to predict death, and AUC = 0.61 to predict the need for admission to the intensive care unit (ICU). AdaBoost had similar performance with ElasticNet in predicting the need for admission to ICU. Analysis of the key radiomic metrics that drive model prediction and performance showed the importance of first-order texture metrics compared to other radiomics panel metrics. Using a Venn-diagram analysis, two first-order texture metrics and one second-order texture metric that consistently played an important role in driving model performance in all three outcome predictions were identified.

Conclusions: Considering the quantitative nature and reliability of radiomic metrics, they can be used prospectively as prognostic markers to individualize treatment plans for COVID-19 patients and also assist with healthcare resource management.

Advances in knowledge: We report on the performance of CXR-based imaging metrics extracted from RT-PCR positive COVID-19 patients at admission to develop machine-learning algorithms for predicting the need for ICU, the need for intubation, and mortality, respectively.

Figure 1.

A typical radiomics workflow showing its four main Stages 1. Image acquisition 2. Segmentation and/or ROI marking (highlighted in red) 3. Feature extraction and finally 4. Statistical analysis. The two green axes divide the image plane into four quadrants. We use 237 radiomic metrics across seven different texture families for this study.

Figure 2.

Area under the curve (AUC) plotted along the x-axis for the three classifiers (i.e., Ada Boost, Elastic Net and Random Forest) considered in the study for predicting the need for ICU (A), need for intubation (B) and mortality (C). Of the three classifiers, Ada Boost shows the best performance for predicting the need for intubation and mortality with an AUC of 0.72 and 0.71, respectively. It has similar performance with ElasticNet in predicting ICU with an AUC of 0.61.

Figure 3.

Variable (radiomic metric) of importance is plotted along the y-axis for the Adaboost model across the three outcome predictions i.e., need for ICU, need of intubation and death, respectively based on ranking of radiomic metrics within a rigorous LOO cross-validation procedure. ‘Frequency’ defined as the number of times each variable made to the top 10 variable of importance list during 10-fold cross-validation is plotted along the x-axis.

Figure 4.

Venn diagram showing the overlap in radiomic metrics between the three prediction models. Of the 3-common overlapping radiomic features across the three prediction models, two belong to the first-order texture metrics: Histogram analysis. The first one, MeanAbsoluteDeviation measures the average distance between each data value and the mean. The metric provides a quantification of the “spread” of the values in a data set. The other histogram metric, entropy help quantifies the information contained within the dataset. Lastly the GLSZM metric: ZoneSizeEntropy evaluates entropy (or uniformity) in the distribution of groups of connected voxels with the same discretized intensity.