Adaptive radiotherapy (ART) is commonly used to mitigate effects of anatomical change during head and neck (H&N) radiotherapy. The process of identifying patients for ART can be subjective and resource-intensive. This feasibility project aims to design and validate a pipeline to automate the process and use it to assess the current clinical pathway for H&N treatments.

The pipeline analysed patients’ on-set cone-beam CT (CBCT) scans to identify inter-fractional anatomical changes. CBCTs were converted into synthetic CTs, contours were automatically generated, and the original plan was recomputed. Each synthetic CT was evaluated against a set of dosimetric goals, with failed goals causing an ART recommendation.

To validate pipeline performance, a ‘gold standard’ was synthesised by recomputing patients’ original plans on a rescan-CT acquired during treatment and identifying failed clinical goals. The pipeline sensitivity and specificity compared to this ‘gold standard’ were calculated for 12 ART patients. The pipeline was then run on a cohort of 12 ART and 14 non-ART patients, and its sensitivity and specificity were instead calculated against the clinical decision made.

The pipeline showed good agreement with the synthesised ‘gold standard’ with an optimum sensitivity of 0·83 and specificity of 0·67. When run over a cohort containing both ART and non-ART patients and assessed against the subjective clinical decision made, the pipeline showed no predictive power (sensitivity: 0·58, specificity: 0·47).

Good agreement with the ‘gold standard’ gives confidence in pipeline performance and disagreement with clinical decisions implies implementation could help standardise the current clinical pathway.

This project developed and validated an automated pipeline for prostate treatments to accurately determine which patients could benefit from adaptive radiotherapy (ART) using synthetic CTs (sCTs) generated from on-treatment cone-beam CT (CBCT) images.

The automated pipeline converted CBCTs to sCTs utilising deep-learning, for accurate dose recalculation. Deformable image registration mapped contours from the planning CT to the sCT, with the treatment plan recalculated. A pass/fail assessment used relevant clinical goals. A fail threshold indicated ART was required. All acquired CBCTs (230 sCTs) for 31 patients (6 who had ART) were assessed for pipeline accuracy and clinical viability, comparing clinical outcomes to pipeline outcomes.

The pipeline distinguished patients requiring ART; 74·4% of sCTs for ART patients were red (failure) results, compared to 6·4% of non-ART sCTs. The receiver operator characteristic area under curve was 0·98, demonstrating high performance. The automated pipeline was statistically significantly (p < 0·05) quicker than the current clinical assessment methods (182·5s and 556·4s, respectively), and deformed contour accuracy was acceptable, with 96·6% of deformed clinical target volumes (CTVs) clinically acceptable.

The automated pipeline identified patients who required ART with high accuracy while reducing time and resource requirements. This could reduce departmental workload and increase efficiency and personalisation of patient treatments. Further work aims to apply the pipeline to other treatment sites and investigate its potential for taking into account dose accumulation.

Advanced laryngeal cancers are clinically complex; there is a paucity of modern decision-making models to guide tumour-specific management. This pilot study aims to identify computed tomography-based radiomic features that may predict survival and enhance prognostication.

Pre-biopsy, contrast-enhanced computed tomography scans were assembled from a retrospective cohort (n = 72) with advanced laryngeal cancers (T3 and T4). The LIFEx software was used for radiomic feature extraction. Two features: shape compacity (irregularity of tumour volume) and grey-level zone length matrix – grey-level non-uniformity (tumour heterogeneity) were selected via least absolute shrinkage and selection operator-based Cox regression and explored for prognostic potential.

A greater shape compacity (hazard ratio 2.89) and grey-level zone length matrix – grey-level non-uniformity (hazard ratio 1.64) were significantly associated with worse 5-year disease-specific survival (p < 0.05). Cox regression models yielded a superior C-index when incorporating radiomic features (0.759) versus clinicopathological variables alone (0.655).

Two radiomic features were identified as independent prognostic biomarkers. A multi-centre prospective study is necessary for further exploration. Integrated radiomic models may refine the treatment of advanced laryngeal cancers.