Impact of Wavelet Kernels on Predictive Capability of Radiomic Features: A Case Study on COVID-19 Chest X-ray Images

Abstract

Radiomic analysis allows for the detection of imaging biomarkers supporting decision-making processes in clinical environments, from diagnosis to prognosis. Frequently, the original set of radiomic features is augmented by considering high-level features, such as wavelet transforms. However, several wavelets families (so called kernels) are able to generate different multi-resolution representations of the original image, and which of them produces more salient images is not yet clear. In this study, an in-depth analysis is performed by comparing different wavelet kernels and by evaluating their impact on predictive capabilities of radiomic models. A dataset composed of 1589 chest X-ray images was used for COVID-19 prognosis prediction as a case study. Random forest, support vector machine, and XGBoost were trained (on a subset of 1103 images) after a rigorous feature selection strategy to build-up the predictive models. Next, to evaluate the models generalization capability on unseen data, a test phase was performed (on a subset of 486 images). The experimental findings showed that Bior1.5, Coif1, Haar, and Sym2 kernels guarantee better and similar performance for all three machine learning models considered. Support vector machine and random forest showed comparable performance, and they were better than XGBoost. Additionally, random forest proved to be the most stable model, ensuring an appropriate balance between sensitivity and specificity.

Shallow and deep learning classifiers in medical image analysis

“…In the case of image analysis, the classifiers discussed in the previous section require that images or ROIs are described through features: these can be handcrafted features and represent the relationships between the gray levels, texture, or shape of a ROI [ 8 , 9 ]. It is also possible to extract higher-level handcrafted features such as wavelet features, which showed remarkably interesting results in several tasks [ 50 ]. CNNs, conversely, include feature extraction in their workflow: given an input image or ROI, they extract the most informative features and then exploit these features for classification using the abovementioned MLP (often referred as “dense layers”) [ 10 ].…”

Improving the Reproducibility of Computed Tomography Radiomic Features Using an Enhanced Hierarchical Feature Synthesis Network

“…In contrast, for wavelettransformed images, it was adjusted to 10 [33]. Coiflets 1 was selected for the wavelet analysis, which is Pyradiomics default and has been utilized by several studies [38][39][40].…”

Explainable Machine-Learning Models for COVID-19 Prognosis Prediction Using Clinical, Laboratory and Radiomic Features

“…Then the same features were extracted considering Laplacian of Gaussian (LoG) and Wavelets filtered images. For LoG filtering three different values of σ were considered (σ ∈ {1, 3, 5}), collecting 279 features (279 = 93 × 3); for Wavelets transform, the Haar kernel [47] and two decomposition levels (levels ∈ {1, 2}) were considered, obtaining 651 features (651 = 93 × 7). Finally, 930 features were extracted from the filtered images.…”