Innovations in thoracic imaging: CT, radiomics, AI and x-ray velocimetry

Abstract

In recent years, pulmonary imaging has seen enormous progress, with the introduction, validation and implementation of new hardware and software. There is a general trend from mere visual evaluation of radiological images to quantification of abnormalities and biomarkers, and assessment of 'non visual' markers that contribute to establishing diagnosis or prognosis. Important catalysts to these developments in thoracic imaging include new indications (like computed tomography [CT] lung cancer screening) and the COVID-19 pandemic. This review focuses on developments in CT, radiomics, artificial intelligence (AI) and x-ray velocimetry for imaging of the lungs. Recent developments in CT include the potential for ultra-low-dose CT imaging for lung nodules, and the advent of a new generation of CT systems based on photon-counting detector technology. Radiomics has demonstrated potential towards predictive and prognostic tasks particularly in lung cancer, previously not achievable by visual inspection by radiologists, exploiting high dimensional patterns (mostly texture related) on medical imaging data. Deep learning technology has revolutionized the field of AI and as a result, performance of AI algorithms is approaching human performance for an increasing number of specific tasks. X-ray velocimetry integrates x-ray (fluoroscopic) imaging with unique image processing to produce quantitative four dimensional measurement of lung tissue motion, and accurate calculations of lung ventilation.

Keywords: computed tomography; deep learning; lung cancer; lung nodules; machine learning; radiomics; x-ray velocimetry.

FIGURE 1

Standard dose CT (HRCT) and ultra‐low‐dose CT (ULD) in the same patient (from the cohort described in Reference 10). (A) Shows a standard reconstructed HRCT image (filtered back projection). (B) Shows a cropped view of the standard reconstructed HRCT image. (C) Shows a standard reconstructed ULD CT image, with elevated image noise. (D) Shows an ULD CT image based on deep learning reconstruction. (E) Shows an ULD CT image based on iterative reconstruction. (D) and (E) show less image noise, more similar to standard dose (HRCT) image

FIGURE 2

Consecutive steps in a typical radiomics workflow (adapted from reference 30)

FIGURE 3

Example of the output of an artificial intelligence (AI) algorithm for lung localization and nodule detection on chest computed tomography (CT) imaging. The lung localization algorithm takes a slice of a CT scan as input, and produces a bounding box around the left lung and the right lung. The lung nodule detection algorithm takes a CT scan as input, and produces bounding boxes around detected lung nodules, which can be presented to radiologists as an aid for the detection of nodules in chest CT images

FIGURE 4

Preclinical experiments validating the accuracy and validity of lung volume measures using XV technology, with (A) and (B) showing bench‐top measurements of fluid flow validated against computer modelling; and (C) and (D) showing in vivo measurements of ventilation in rabbit lungs validated against plethysmography. (A) Reconstructed 3D blood velocity flow fields measured using XV. For clarity only half the sample is plotted, with reduced vector resolution in all dimensions. Vector colours represent velocity magnitude and are validated against computational models of the flow field. (B) CT XV reconstruction of flow field through helical geometry. A section of the result has been rendered as transparent for visualization of the flow. The results indicate the ability of CT XV to simultaneously measure the 3D structure and velocity of flow through complex geometries. (C) In vivo measurements of ventilation in rabbit lungs with validation of integrated divergence (volume) measurements from XV technology against volume measures from plethysmography. A scatter plot shows strong correlation between two quantities. (D) Time series of lung volume co‐plotted with divergence demonstrated a direct link between divergence and tissue expansion

FIGURE 5

(A) Distribution of regional lung ventilation during XV scanning is shown using a colour scale where red represents underventilation, green represents average ventilation and blue represents hyperventilation relative to the mean regional lung volume expansion. The visualization maps show a mid‐coronal slice and axial slices from the upper, middle and lower zones at peak inspiration. (B) Lobe‐wise XV analysis performed using an automated anatomy‐based segmentation