Investigating the use of machine learning to generate ventilation images from CT scans

Abstract

Background Radiotherapy treatment planning incorporating ventilation imaging can reduce the incidence of radiation‐induced lung injury. The gold‐standard of ventilation imaging, using nuclear medicine, has limitations with respect to availability and cost. Purpose An alternative type of ventilation imaging to nuclear medicine uses 4DCT (or breath‐hold CT [BHCT] pair) with deformable image registration (DIR) and a ventilation metric to produce a CT ventilation image (CTVI). The purpose of this study is to investigate the application of machine learning as an alternative to DIR‐based methods when producing CTVIs. Methods A patient dataset of 15 inhale and exhale BHCTs and Galligas PET ventilation images were used to train and test a 2D U‐Net style convolutional neural network. The neural network established relationships between axial input BHCT image pairs and axial labeled Galligas PET images and was evaluated using eightfold cross‐validation. Once trained, the neural network could produce a CTVI from an input BHCT image pair. The CTVIs produced by the neural network were qualitatively assessed visually and quantitatively compared to a Galligas PET ventilation image using a Spearman correlation and Dice similarity coefficient (DSC). The DSC measured the spatial overlap between three segmented equal lung volumes by ventilation (high, medium, and low functioning lung [LFL]). Results The mean Spearman correlation between the CTVIs and the Galligas PET ventilation images was 0.58 ± 0.14. The mean DSC over high, medium, and LFL between the CTVIs and Galligas PET ventilation images was 0.55 ± 0.06. Visually, a systematic overprediction of ventilation within the lung was observed in the CTVIs with respect to the Galligas PET ventilation images, with jagged regions of ventilation in the sagittal and coronal planes. Conclusions A convolutional neural network was developed that could produce a CTVI from a BHCT image pair, which was then compared with a Galligas PET ventilation image. The performance of this machine learning method was comparable to previous benchmark studies investigating a DIR‐based CTVI, warranting future development, and investigation of applying machine learning to a CTVI.

Respiratory Invariant Textures From Static Computed Tomography Scans for Explainable Lung Function Characterization

“…Deformable image registration (DIR)-based CTVI/PI methods can mathematically derive function distribution via registered biphasic breath-hold CT images or biphasic image pairs from 4-dimensional computed tomography (4DCT) scans. Few artificial intelligence (AI)-based CTVI/PI models 5,[17][18][19] have also been proposed to directly generate NM-like lung function images through the end-to-end convolutional neural network. However, DIR-based CTVI/PI has been demonstrated to be sensitive to the input image quality and the DIR algorithm.…”

“…Third, the mapping of feature distribution also offered greater explainability than the AIbased method. Instead of the black box in the published AI-based methods, 5,[17][18][19] our method provides a definitive mathematical algorithm with a sliding window technique for function distribution mapping. From a mathematical perspective, the GLDM dependence nonuniformity values measure the homogeneity among dependencies of the image intensity, which are proportionally related to the coarseness of the image.…”

PixelPrint4D: A 3D printing method of fabricating patient-specific deformable CT phantoms for respiratory motion applications

“…These allow for the modulation of the radiation beam to better target the expected trajectory of tumor motion. Newer technologies have also been developed, including deformable image registration (DIR), deep learning-based lung tumor prediction models 6,7 , motion artifact reduction algorithms 8,9 , and CT ventilation imaging 10,11 .…”

A hybrid model‐ and deep learning‐based framework for functional lung image synthesis from multi‐inflation CT and hyperpolarized gas MRI

“…Ten-fold cross-validation was used, achieving an average DSC across all folds of 0.83 for high-functional lung regions, 0.61 for mediumfunctional lung regions, and 0.73 for low-functional lung regions. Subsequently, Grover et al 25 investigated the utility of CNNs for synthesizing Galligas PET ventilation images, demonstrating a mean Spearman's correlation of 0.58 and a mean DSC for high, medium, and low functional regions of 0.55. However, SPECT and PET have significantly longer acquisition times compared to CT imaging which facilitates acquisition within a single breath.…”

“…Subsequently, Grover et al. 25 investigated the utility of CNNs for synthesizing Galligas PET ventilation images, demonstrating a mean Spearman's correlation of 0.58 and a mean DSC for high, medium, and low functional regions of 0.55. However, SPECT and PET have significantly longer acquisition times compared to CT imaging which facilitates acquisition within a single breath.…”