Gross failure rates and failure modes for a commercial AI-based auto-segmentation algorithm in head and neck cancer patients.

PURPOSE: Artificial intelligence (AI) based commercial software can be used to automatically delineate organs at risk (OAR), with potential for efficiency savings in the radiotherapy treatment planning pathway, and reduction of inter- and intra-observer variability. There has been little research investigating gross failure rates and failure modes of such systems. METHOD: 50 head and neck (H&N) patient data sets with "gold standard" contours were compared to AI-generated contours to produce expected mean and standard deviation values for the Dice Similarity Coefficient (DSC), for four common H&N OARs (brainstem, mandible, left and right parotid). An AI-based commercial system was applied to 500 H&N patients. AI-generated contours were compared to manual contours, outlined by an expert human, and a gross failure was set at three standard deviations below the expected mean DSC. Failures were inspected to assess reason for failure of the AI-based system with failures relating to suboptimal manual contouring censored. True failures were classified into 4 sub-types (setup position, anatomy, image artefacts and unknown). RESULTS: There were 24 true failures of the AI-based commercial software, a gross failure rate of 1.2%. Fifteen failures were due to patient anatomy, four were due to dental image artefacts, three were due to patient position and two were unknown. True failure rates by OAR were 0.4% (brainstem), 2.2% (mandible), 1.4% (left parotid) and 0.8% (right parotid). CONCLUSION: True failures of the AI-based system were predominantly associated with a non-standard element within the CT scan. It is likely that these non-standard elements were the reason for the gross failure, and suggests that patient datasets used to train the AI model did not contain sufficient heterogeneity of data. Regardless of the reasons for failure, the true failure rate for the AI-based system in the H&N region for the OARs investigated was low (∼1%).

Dose intensified hypofractionated intensity-modulated radiotherapy with synchronous cetuximab for intermediate stage head and neck squamous cell carcinoma.

BACKGROUND: For stage II and III head and neck squamous cell carcinoma (HNSCC) treated with radiotherapy alone, loco-regional recurrence is the main cause of treatment failure. Strategies to improve loco-regional control should not be at the expense of increased late normal tissue toxicity. We investigated dose-intensified hypofractionated intensity-modulated radiotherapy (IMRT) with synchronous cetuximab. MATERIAL AND METHODS: In a phase I/II trial, 27 patients with stage III or high risk stage II HNSCC were recruited. They received three dose level simultaneous integrated boost IMRT, 62.5 Gy in 25 daily fractions to planning target volume one over five weeks with synchronous cetuximab. The primary endpoint was acute toxicity. Secondary endpoints included: late toxicity and quality of life; loco-regional control, cause-specific and overall survival. RESULTS: Radiotherapy was completed by 26/27 patients; for one (4%) the final fraction was omitted due to skin toxicity. All cycles of cetuximab were received by 23/27 patients. Grade 3 acute toxicities included: pain (81%), oral mucositis (78%) and dysphagia (41%). There were few grade 3 physician-recorded late toxicities, including: pain (11%), problems with teeth (8%) and weight loss (4%). At 12 months, only one (4%) patient required a feeding tube, inserted prior to treatment due to dysphagia. The maximal/peak rates of patient-reported late toxicities included: severe pain (11%), any dry mouth (89%) and swallowing dysfunction that required a soft/liquid diet (23%). At 12 months, all quality of life and most symptoms mean scores had resolved to baseline or were only a little worse; dry mouth, sticky saliva and dentition scores remained very much worse. At a median follow-up of 47 months, there were five (18.5%) loco-regional recurrences and the overall cause-specific survival was 79% (95% CI 53-92). CONCLUSIONS: This regimen is safe with acceptable acute toxicity, low rates of late toxicity and impact on quality of life at 12 months following treatment. Further evaluation is recommended.

Evaluation of the impact of dental artefacts on intensity-modulated radiotherapy planning for the head and neck.

BACKGROUND AND PURPOSE: High density materials create severe artefacts in the computed tomography (CT) scans used for radiotherapy dose calculations. Increased use of intensity-modulated radiotherapy (IMRT) to treat oropharyngeal cancers raises concerns over the accuracy of the resulting dose calculation. This work quantifies their impact and evaluates a simple corrective technique. MATERIALS AND METHODS: Fifteen oropharyngeal patients with severe artefacts were retrospectively planned with IMRT using two different CT/density look-up tables. Each plan was recalculated using a corrected CT dataset to evaluate the dose distribution delivered to the patient. Plan quality in the absence of dental artefacts was similarly assessed. A range of dosimetric and radiobiological parameters were compared pre- and post-correction. RESULTS: Plans using a standard CT/density look-up table (density 1.8 g/cm(3)) revealed inconsistent inter-patient errors, mostly within clinical acceptance, although potentially significantly reducing target coverage for individual patients. Using an extended CT/density look-up table (density 10.0 g/cm(3)) greatly reduced the errors for 13/15 patients. In 2/15 patients with residual errors the CTV extended into the severely affected region and could be corrected by applying a simple manual correction. CONCLUSIONS: Use of an extended CT/density look-up table together with a simple manual bulk density correction reduces the impact of dental artefacts on head and neck IMRT planning to acceptable levels.