CBCT correction using a cycle-consistent generative adversarial network and unpaired training to enable photon and proton dose calculation

Abstract

In presence of inter-fractional anatomical changes, clinical benefits are anticipated from image-guided adaptive radiotherapy. Nowadays, cone-beam CT (CBCT) imaging is mostly utilized during pre-treatment imaging for position verification. Due to various artifacts, image quality is typically not sufficient for photon or proton dose calculation, thus demanding accurate CBCT correction, as potentially provided by deep learning techniques. This work aimed at investigating the feasibility of utilizing a cycle-consistent generative adversarial network (cycleGAN) for prostate CBCT correction using unpaired training. Thirty-three patients were included. The network was trained to translate uncorrected, original CBCT images (CBCTorg) into planning CT equivalent images (CBCTcycleGAN). HU accuracy was determined by comparison to a previously validated CBCT correction technique (CBCTcor). Dosimetric accuracy was inferred for volumetric-modulated arc photon therapy (VMAT) and opposing single-field uniform dose (OSFUD) proton plans, optimized on CBCTcor and recalculated on CBCTcycleGAN. Single-sided SFUD proton plans were utilized to assess proton range accuracy. The mean HU error of CBCTcycleGAN with respect to CBCTcor decreased from 24 HU for CBCTorg to −6 HU. Dose calculation accuracy was high for VMAT, with average pass-rates of 100%/89% for a 2%/1% dose difference criterion. For proton OSFUD plans, the average pass-rate for a 2% dose difference criterion was 80%. Using a (2%, 2 mm) gamma criterion, the pass-rate was 96%. 93% of all analyzed SFUD profiles had a range agreement better than 3 mm. CBCT correction time was reduced from 6–10 min for CBCTcor to 10 s for CBCTcycleGAN. Our study demonstrated the feasibility of utilizing a cycleGAN for CBCT correction, achieving high dose calculation accuracy for VMAT. For proton therapy, further improvements may be required. Due to unpaired training, the approach does not rely on anatomically consistent training data or potentially inaccurate deformable image registration. The substantial speed-up for CBCT correction renders the method particularly interesting for adaptive radiotherapy.

Comparison of CBCT‐based dose calculation methods in head and neck cancer radiotherapy: from Hounsfield unit to density calibration curve to deep learning

“…Studies using DL for pCT generation from CBCT are scarce for brain, 55 H&N, [13][14][15] pancreas, 38 or prostate cancer. [35][36][37]55,56 The studies involved an imaging analysis (pCT vs. reference CT), but only half of them evaluated the dose accuracy. Among the three H&N studies using DL for pCT generation from CBCT, [13][14][15] one involved training a U-Net neural network on 50 coregistered CBCT/CT images and performing a test based on data from 10 patients.…”

“…32 Some studies have recently proposed DLMs for pCT generation from CBCT, mainly for scatter correction. 33,34 Other studies proposed the DLM for pCT generation from CBCT in prostate, [35][36][37] pancreas, 38 and H&N, [13][14][15] for dose calculation. H&N studies have been performed using either U-Net or cycleGAN architectures to generate pCT from CBCT.…”

Cone‐beam CT‐derived relative stopping power map generation via deep learning for proton radiotherapy

“…The calculation of MAE and GD loss is introduced in our previous studies. 35,40 MI is a measure of mutual dependency between two probability distributions…”

“…33,34 Kurz et al recently published a deep-learning-based CBCT correction method, which they validated on photon and proton treatments for prostate cancer. 35 The goal of these studies has been to synthesize a CT image that can be used for treatment planning based on the patient's current setup on the treatment table. In this work, we propose to extend a cycle-consistent generative adversarial network (cycle-GAN) from learning a CBCT-to-CT image transformation 36 to learning a CBCT-to-RSP map transformation.…”

Overview of artificial intelligence-based applications in radiotherapy: Recommendations for implementation and quality assurance

“…Thus far, the most practical and common approach is deformable image registration to map planning CT HU to the (daily) treatment image. Recently, it has been shown that CNNs (DL) provide promising methods for synthetic CT generation based on CBCT or MRI images [86,[94][95][96][97][98][99].…”