Deep learning enables MV-based real-time image guided radiation therapy for prostate cancer patients

Using MV images for real-time image guided radiation therapy (IGRT) is ideal as it does not require additional imaging equipment, adds no additional imaging dose and provides motion data in the treatment beam frame of reference. However, accurate tracking using MV images is challenging due to low contrast and modulated fields. Here, a novel real-time marker tracking system based on a convolutional neural network (CNN) classifier was developed and evaluated on retrospectively acquired patient data for MV-based IGRT for prostate cancer patients. 

MV images, acquired from 29 VMAT prostate cancer patients treated in a multi-institutional clinical trial, were used to train and evaluate a CNN-based marker tracking system. The CNN was trained using labelled MV images from 9 prostate cancer patients (35 fractions) with implanted markers. CNN performance was evaluated on an independent cohort of unseen MV images from 20 patients (78 fractions), using a Precision-Recall curve (PRC), area under the PRC plot (AUC) and sensitivity and specificity. The accuracy of the tracking system was evaluated on the same unseen dataset and quantified by calculating mean absolute (± 1 SD) and [1st, 99th] percentiles of the geometric tracking error in treatment beam co-ordinates using manual identification as the ground truth. 

The CNN had an AUC of 0.99, sensitivity of 98.31% and specificity of 99.87%. The mean absolute geometric tracking error was 0.30 ± 0.27 and 0.35 ± 0.31 mm in the lateral and superior-inferior directions of the MV images, respectively. The [1st, 99th] percentiles of the error were [-1.03, 0.90] and [-1.12, 1.12] mm in the lateral and SI directions, respectively.

The high classification performance on unseen MV images demonstrates the CNN can successfully identify implanted prostate markers. Furthermore, the sub-millimetre accuracy and precision of the marker tracking system demonstrates potential for adaptation to real-time applications.

Dose-based optimisation for multi-leaf collimator tracking during radiation therapy

Motion in the patient anatomy causes a reduction in dose delivered to the target, while increasing dose to healthy tissue. Multi-leaf collimator (MLC) tracking has been clinically implemented to adapt dose delivery to account for intrafraction motion. Current methods shift the planned MLC aperture in the direction of motion, then optimise the new aperture based on the difference in fluence. The drawback of these methods is that 3D dose, a function of patient anatomy and MLC aperture sequence, is not properly accounted for. To overcome the drawback of current fluence-based methods, we have developed and investigated real-time adaptive MLC tracking based on dose optimisation. A novel MLC tracking algorithm, dose optimisation, has been developed which accounts for the moving patient anatomy by optimising the MLC based on the dose delivered during treatment, simulated using a simplified dose calculation algorithm. The MLC tracking with dose optimisation method was applied in silico to a prostate cancer VMAT treatment dataset with observed intrafraction motion. Its performance was compared to MLC tracking with fluence optimisation and, as a baseline, without MLC tracking. To quantitatively assess performance, we computed the dose error and 3D γ failure rate (2 mm/2%) for each fraction and method. Dose optimisation achieved a γ failure rate of (4.7 ± 1.2)% (mean and standard deviation) over all fractions, which was significantly lower than fluence optimisation (7.5 ± 2.9)% (Wilcoxon sign-rank test p < 0.01). Without MLC tracking, a γ failure rate of (15.3 ± 12.9)% was achieved. By considering the accumulation of dose in the moving anatomy during treatment, dose optimisation is able to optimise the aperture to actively target regions of underdose while avoiding overdose.

Real-Time Image Guided Ablative Prostate Cancer Radiation Therapy: Results From the TROG 15.01 SPARK Trial

PURPOSE

Kilovoltage intrafraction monitoring (KIM) is a novel software platform implemented on standard radiation therapy systems and enabling real-time image guided radiation therapy (IGRT). In a multi-institutional prospective trial, we investigated whether real-time IGRT improved the accuracy of the dose patients with prostate cancer received during radiation therapy.

METHODS AND MATERIALS

Forty-eight patients with prostate cancer were treated with KIM-guided SABR with 36.25 Gy in 5 fractions. During KIM-guided treatment, the prostate motion was corrected for by either beam gating with couch shifts or multileaf collimator tracking. A dose reconstruction method was used to evaluate the dose delivered to the target and organs at risk with and without real-time IGRT. Primary outcome was the effect of real-time IGRT on dose distributions. Secondary outcomes included patient-reported outcomes and toxicity.

RESULTS

Motion correction occurred in ≥1 treatment for 88% of patients (42 of 48) and 51% of treatments (121 of 235). With real-time IGRT, no treatments had prostate clinical target volume (CTV) D98% dose 5% less than planned. Without real-time IGRT, 13 treatments (5.5%) had prostate CTV D98% doses 5% less than planned. The prostate CTV D98% dose with real-time IGRT was closer to the plan by an average of 1.0% (range, -2.8% to 20.3%). Patient outcomes showed no change in the 12-month patient-reported outcomes compared with baseline and no grade ≥3 genitourinary or gastrointestinal toxicities.

CONCLUSIONS

Real-time IGRT is clinically effective for prostate cancer SABR.