Fig. 4

QuantImage v2 (QI2) feature explorer and visualization. The Visualization tab of the QI2 Feature Explorer provides feature selection and visualization functionalities. A tree-like filter mechanism allows selecting specific features based on the imaging modality (e.g., PET, CT) and ROI (here: “GTV L”) from which they were derived and grouped by feature class (e.g., “texture”). A heatmap visualizes the values of the selected features for all chosen patients ordered by the outcome measure. Automated feature selection and ranking options can be applied to refine this view. The three selected features displayed here (CT kurtosis, CT skewness derived from Laplacian-of-Gaussian filtered images, and PET SUVmax) are highly predictive of the presence of pulmonary lymphangitic carcinomatosis (average area under the curve of 0.94 and 0.88 from cross-validation and on test set, respectively). PET, Positron emission tomography; CT, Computed tomography; ROI, Region of interest; SUV, Standardized uptake value