This is the authors' abstract. We don't add key points for this paper.
Purpose Lung tumor tracking with CyberKnife sometimes misrecognizes tumor location when there are overlapping bony structures in the search area near the tumor. This study aimed to develop a bone suppression technique for kV-X-ray imaging using machine learning. Methods CT images were created with or without bony structures using a 4D extended cardiac-torso phantom (XCAT phantom) in 56 cases. Subsequently, X-ray projection images in the left and right oblique 45-degree directions from these images were then divided into nine squares for 1120 images. Images with bone were input into six general adversarial network (GAN) and the bone suppressed images on the XCAT phantom (BSIphantom) were created. They were then compared to images without bone using the structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR), frechet inception distance (FID). Next, the bone suppressed image (BSIpatient) were created from 1000 non-simulated treatment images of the actual patient. Zero means normalized cross correlation (ZNCC) by template matching between each of treatment images and BSIpatient were calculated. Results BSIphantom were calculated the highest SSIM and PSNR of 0.96 ± 0.02 and 36.93 ± 3.93, respectively, compared to images without bone shadows in the test data. Five out of six algorithms significantly improved ZNCC, showing a higher ZNCC than the actual treatment images. FID evaluation was showed that the SimDCL was the most promising model with 68.93. Conclusions The proposed GAN-based bone suppression image processing technique improved image recognition and enabled highly accurate dynamic tracking irradiation.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Mochizuki et al. (2023) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: