First experience of autonomous, un-supervised treatment planning integrated in adaptive MR-guided radiotherapy and delivered to a patient with prostate cancer

The commissioning of a three-dimensional treatment planning system requires comparisons of measured and calculated dose distributions. Techniques have been developed to facilitate quantitative comparisons, including superimposed isodoses, dose-difference, and distance-to-agreement (DTA) distributions. The criterion for acceptable calculation performance is generally defined as a tolerance of the dose and DTA in regions of low and high dose gradients, respectively. The dose difference and DTA distributions complement each other in their useful regions. A composite distribution has recently been developed that presents the dose difference in regions that fail both dose-difference and DTA comparison criteria. Although the composite distribution identifies locations where the calculation fails the preselected criteria, no numerical quality measure is provided for display or analysis. A technique is developed to unify dose distribution comparisons using the acceptance criteria. The measure of acceptability is the multidimensional distance between the measurement and calculation points in both the dose and the physical distance, scaled as a fraction of the acceptance criteria. In a space composed of dose and spatial coordinates, the acceptance criteria form an ellipsoid surface, the major axis scales of which are determined by individual acceptance criteria and the center of which is located at the measurement point in question. When the calculated dose distribution surface passes through the ellipsoid, the calculation passes the acceptance test for the measurement point. The minimum radial distance between the measurement point and the calculation points (expressed as a surface in the dose-distance space) is termed the gamma index. Regions where gamma > 1 correspond to locations where the calculation does not meet the acceptance criteria. The determination of gamma throughout the measured dose distribution provides a presentation that quantitatively indicates the calculation accuracy. Examples of a 6 MV beam penumbra are used to illustrate the gamma index.

The complexity and rise of data in healthcare means that artificial intelligence (AI) will increasingly be applied within the field. Several types of AI are already being employed by payers and providers of care, and life sciences companies. The key categories of applications involve diagnosis and treatment recommendations, patient engagement and adherence, and administrative activities. Although there are many instances in which AI can perform healthcare tasks as well or better than humans, implementation factors will prevent large-scale automation of healthcare professional jobs for a considerable period. Ethical issues in the application of AI to healthcare are also discussed.

Summary Background Prostate cancer might have high radiation-fraction sensitivity that would give a therapeutic advantage to hypofractionated treatment. We present a pre-planned analysis of the efficacy and side-effects of a randomised trial comparing conventional and hypofractionated radiotherapy after 5 years follow-up. Methods CHHiP is a randomised, phase 3, non-inferiority trial that recruited men with localised prostate cancer (pT1b–T3aN0M0). Patients were randomly assigned (1:1:1) to conventional (74 Gy delivered in 37 fractions over 7·4 weeks) or one of two hypofractionated schedules (60 Gy in 20 fractions over 4 weeks or 57 Gy in 19 fractions over 3·8 weeks) all delivered with intensity-modulated techniques. Most patients were given radiotherapy with 3–6 months of neoadjuvant and concurrent androgen suppression. Randomisation was by computer-generated random permuted blocks, stratified by National Comprehensive Cancer Network (NCCN) risk group and radiotherapy treatment centre, and treatment allocation was not masked. The primary endpoint was time to biochemical or clinical failure; the critical hazard ratio (HR) for non-inferiority was 1·208. Analysis was by intention to treat. Long-term follow-up continues. The CHHiP trial is registered as an International Standard Randomised Controlled Trial, number ISRCTN97182923. Findings Between Oct 18, 2002, and June 17, 2011, 3216 men were enrolled from 71 centres and randomly assigned (74 Gy group, 1065 patients; 60 Gy group, 1074 patients; 57 Gy group, 1077 patients). Median follow-up was 62·4 months (IQR 53·9–77·0). The proportion of patients who were biochemical or clinical failure free at 5 years was 88·3% (95% CI 86·0–90·2) in the 74 Gy group, 90·6% (88·5–92·3) in the 60 Gy group, and 85·9% (83·4–88·0) in the 57 Gy group. 60 Gy was non-inferior to 74 Gy (HR 0·84 [90% CI 0·68–1·03], pNI=0·0018) but non-inferiority could not be claimed for 57 Gy compared with 74 Gy (HR 1·20 [0·99–1·46], pNI=0·48). Long-term side-effects were similar in the hypofractionated groups compared with the conventional group. There were no significant differences in either the proportion or cumulative incidence of side-effects 5 years after treatment using three clinician-reported as well as patient-reported outcome measures. The estimated cumulative 5 year incidence of Radiation Therapy Oncology Group (RTOG) grade 2 or worse bowel and bladder adverse events was 13·7% (111 events) and 9·1% (66 events) in the 74 Gy group, 11·9% (105 events) and 11·7% (88 events) in the 60 Gy group, 11·3% (95 events) and 6·6% (57 events) in the 57 Gy group, respectively. No treatment-related deaths were reported. Interpretation Hypofractionated radiotherapy using 60 Gy in 20 fractions is non-inferior to conventional fractionation using 74 Gy in 37 fractions and is recommended as a new standard of care for external-beam radiotherapy of localised prostate cancer. Funding Cancer Research UK, Department of Health, and the National Institute for Health Research Cancer Research Network.