Improved contrast and noise of megavoltage computed tomography (MVCT) through cycle‐consistent generative machine learning

Abstract

Purpose: Megavoltage computed tomography (MVCT) has been implemented on many radiation therapy treatment machines as a tomographic imaging modality that allows for three-dimensional visualization and localization of patient anatomy. Yet MVCT images exhibit lower contrast and greater noise than its kilovoltage CT (kVCT) counterpart. In this work, we sought to improve these disadvantages of MVCT images through an image-to-image-based machine learning transformation of MVCT and kVCT images. We demonstrated that by learning the style of kVCT images, MVCT images can be converted into high-quality synthetic kVCT (skVCT) images with higher contrast and lower noise, when compared to the original MVCT. Methods: Kilovoltage CT and MVCT images of 120 head and neck (H&N) cancer patients treated on an Accuray TomoHD system were retrospectively analyzed in this study. A cycle-consistent generative adversarial network (CycleGAN) machine learning, a variant of the generative adversarial network (GAN), was used to learn Hounsfield Unit (HU) transformations from MVCT to kVCT images, creating skVCT images. A formal mathematical proof is given describing the interplay between function sensitivity and input noise and how it applies to the error variance of a high-capacity function trained with noisy input data. Finally, we show how skVCT shares distributional similarity to kVCT for various macro-structures found in the body. Results: Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were improved in skVCT images relative to the original MVCT images and were consistent with kVCT images. Specifically, skVCT CNR for muscle-fat, bone-fat, and bone-muscle improved to 14.8 AE 0.4, 122.7 AE 22.6, and 107.9 AE 22.4 compared with 1.6 AE 0.3, 7.6 AE 1.9, and 6.0 AE 1.7, respectively, in the original MVCT images and was more consistent with kVCT CNR values of 15.2 AE 0.8, 124.9 AE 27.0, and 109.7 AE 26.5, respectively. Noise was significantly reduced in skVCT images with SNR values improving by roughly an order of magnitude and consistent with kVCT SNR values. Axial slice mean (S-ME) and mean absolute error (S-MAE) agreement between kVCT and MVCT/skVCT improved, on average, from −16.0 and 109.1 HU to 8.4 and 76.9 HU, respectively. Conclusions: A kVCT-like qualitative aid was generated from input MVCT data through a Cycle-GAN instance. This qualitative aid, skVCT, was robust toward embedded metallic material, dramatically improves HU alignment from MVCT, and appears perceptually similar to kVCT with SNR and CNR values equivalent to that of kVCT images.

Improvement of megavoltage computed tomography image quality for adaptive helical tomotherapy using cycleGAN‐based image synthesis with small datasets

“…Although a non-rigid registration was applied to the planning kVCT images to compensate for the differences in the anatomical structures between MVCT and kVCT, the Recently, skVCT generation from MVCT using conventional cycleGAN has been proposed. 64 This study trained the conventional cycleGAN with kV and MVCT images of 100 head and neck cancer patients and was validated with data from 20 patients. In the validation step, the skVCT from conventional cycleGAN was analyzed at the slice and voxel levels through various quantitative evaluations compared to kVCT.…”

Training of deep cross-modality conversion models with a small dataset, and their application in megavoltage CT to kilovoltage CT conversion

“…The In general, deep learning approaches are believed to be data intensive, and tens to hundreds of thousands of images are typically required for natural image recognition and classification. In the field of modality conversions, certain researchers reported the use of thousands of images [4][5][6][7][8][9]. Nevertheless, the proposed model exhibits a competitive performance with a considerably smaller dataset that consists of merely 256 images.…”

“…We constructed a modality conversion model based on deep learning and applied it to realize MVCT to kVCT conversion to enhance the image quality of MVCT in helical tomotherapy. The proposed model represents several enhancements over the existing approach to realize MVCT to kVCT conversion [9] based on CycleGAN. In particular, our model resulted in a higher structure preservation and reduction in the training data.…”

“…Cone-beam CT (CBCT) to CT conversion has been considered to enhance the image quality of CBCT, and thus, the accuracy of image-guided radiotherapy (IGRT) [6], and to facilitate its application to adaptive radiotherapy (ART) [7,8]. Recently, certain researchers [9] investigated mega-voltage CT (MVCT) to kilo-voltage CT (kVCT) conversion to enhance the image quality of MVCT, which is considerably noisier and has a lower contrast than CBCT. Almost all practical modality conversion methods rely on deep learning (DL).…”

“…However, in modality conversion, many images in the corresponding modalities must be prepared, which is cumbersome for the patients and expensive for the medical staff. For instance, in the abovementioned instances of modality conversions [4][5][6][7][8][9], the training data for DL models ranged from 2795 to 6480 images. Furthermore, a question often arises regarding the number of images necessary to ensure successful training.…”