William Beasley

An automated workflow for patient-specific quality control of contour propagation.

Contour propagation is an essential component of adaptive radiotherapy, but current contour propagation algorithms are not yet sufficiently accurate to be used without manual supervision. Manual review of propagated contours is time-consuming, making routine implementation of real-time adaptive radiotherapy unrealistic. Automated methods of monitoring the performance of contour propagation algorithms are therefore required. We have developed an automated workflow for patient-specific quality control of contour propagation and validated it on a cohort of head and neck patients, on which parotids were outlined by two observers. Two types of error were simulated-mislabelling of contours and introducing noise in the scans before propagation. The ability of the workflow to correctly predict the occurrence of errors was tested, taking both sets of observer contours as ground truth, using receiver operator characteristic analysis. The area under the curve was 0.90 and 0.85 for the observers, indicating good ability to predict the occurrence of errors. This tool could potentially be used to identify propagated contours that are likely to be incorrect, acting as a flag for manual review of these contours. This would make contour propagation more efficient, facilitating the routine implementation of adaptive radiotherapy.

The suitability of common metrics for assessing parotid and larynx autosegmentation accuracy

Contouring structures in the head and neck is time-consuming, and automatic segmentation is an important part of an adaptive radiotherapy workflow. Geometric accuracy of automatic segmentation algorithms has been widely reported, but there is no consensus as to which metrics provide clinically meaningful results. This study investigated whether geometric accuracy (as quantified by several commonly used metrics) was associated with dosimetric differences for the parotid and larynx, comparing automatically generated contours against manually drawn ground truth contours. This enabled the suitability of different commonly used metrics to be assessed for measuring automatic segmentation accuracy of the parotid and larynx. Parotid and larynx structures for 10 head and neck patients were outlined by five clinicians to create ground truth structures. An automatic segmentation algorithm was used to create automatically generated normal structures, which were then used to create volumetric-modulated arc therapy plans. The mean doses to the automatically generated structures were compared with those of the corresponding ground truth structures, and the relative difference in mean dose was calculated for each structure. It was found that this difference did not correlate with the geometric accuracy provided by several metrics, notably the Dice similarity coefficient, which is a commonly used measure of spatial overlap. Surface-based metrics provided stronger correlation and are, therefore, more suitable for assessing automatic segmentation of the parotid and larynx. PACS number(s): 87.57.nm, 87.55.D, 87.55.Qr.Contouring structures in the head and neck is time‐consuming, and automatic segmentation is an important part of an adaptive radiotherapy workflow. Geometric accuracy of automatic segmentation algorithms has been widely reported, but there is no consensus as to which metrics provide clinically meaningful results. This study investigated whether geometric accuracy (as quantified by several commonly used metrics) was associated with dosimetric differences for the parotid and larynx, comparing automatically generated contours against manually drawn ground truth contours. This enabled the suitability of different commonly used metrics to be assessed for measuring automatic segmentation accuracy of the parotid and larynx. Parotid and larynx structures for 10 head and neck patients were outlined by five clinicians to create ground truth structures. An automatic segmentation algorithm was used to create automatically generated normal structures, which were then used to create volumetric‐modulated arc therapy plans. The mean doses to the automatically generated structures were compared with those of the corresponding ground truth structures, and the relative difference in mean dose was calculated for each structure. It was found that this difference did not correlate with the geometric accuracy provided by several metrics, notably the Dice similarity coefficient, which is a commonly used measure of spatial overlap. Surface‐based metrics provided stronger correlation and are, therefore, more suitable for assessing automatic segmentation of the parotid and larynx. PACS number(s): 87.57.nm, 87.55.D, 87.55.Qr

Relative plan robustness of step-and-shoot vs rotational intensity–modulated radiotherapy on repeat computed tomographic simulation for weight loss in head and neck cancer

INTRODUCTIONnInterfractional anatomical alterations may have a differential effect on the dose delivered by step-and-shoot intensity-modulated radiotherapy (IMRT) and volumetric-modulated arc therapy (VMAT). The increased degrees of freedom afforded by rotational delivery may increase plan robustness (measured by change in target volume coverage and doses to organs at risk [OARs]). However, this has not been evaluated for head and neck cancer.nnnMATERIALS AND METHODSnA total of 10 patients who required repeat computed tomography (CT) simulation and replanning during head and neck IMRT were included. Step-and-shoot IMRT and VMAT plans were generated from the original planning scan. The initial and second CT simulation scans were fused and targets/OAR contours transferred, reviewed, and modified. The plans were applied to the second CT scan and doses recalculated without repeat optimization. Differences between step-and-shoot IMRT and VMAT for change in target volume coverage and doses to OARs between first and second CT scans were compared by Wilcoxon signed rank test.nnnRESULTSnThere were clinically relevant dosimetric changes between the first and the second CT scans for both the techniques (reduction in mean D95% for PTV2 and PTV3, Dmin for CTV2 and CTV3, and increased mean doses to the parotid glands). However, there were no significant differences between step-and-shoot IMRT and VMAT for change in any target coverage parameter (including D95% for PTV2 and PTV3 and Dmin for CTV2 and CTV3) or dose to any OARs (including parotid glands) between the first and the second CT scans.nnnCONCLUSIONSnFor patients with head and neck cancer who required replanning mainly due to weight loss, there were no significant differences in plan robustness between step-and-shoot IMRT and VMAT. This information is useful with increased clinical adoption of VMAT.

Learning from every patient treated

Modern precision radiotherapy allows small safety margins and dose escalation. Therefore, biological factors become much more important such as CTV delineation and thresholds for organ at risk tolerance. Our aim is to develop image based data mining for exploring voxel-based dose–response relationships in very large patient cohorts. Large numbers of planning CTs are deformably registered to a reference CT. Registration uncertainties are quantified using organ-at-risk contours, dose distributions are smoothed according to these uncertainties and mapped onto the reference. Next outcome measures are correlated voxel-by-voxel with the dose distributions. The resulting correlation maps are tested for significance using a test statistic, e.g. maximum t-value, using randomization to test for significance. We have applied this methodology in several tumour sites and a great strength of this technique is that it allows discovery of sensitive sub-structures of organs. For example, in lung cancer we demonstrated a relationship of dose to the base of the heart with early mortality (1100 patients); while in head and neck cancer, masseter dose correlated most with post treatment trismus. In prostate cancer, obturator dose relates to PSA control. To understand the results, it is important to study inherent correlations in voxel-wise dose distributions that are related to planning techniques that are often ignored in dose-volume based analyses. We conclude that voxel based dose response relationships can be discovered efficiently using deformable registration and novel statistical techniques and that these complement traditional dose–volume analyses, and are suitable for very large patient cohorts.

Image-based Data Mining to Probe Dosimetric Correlates of Radiation-induced Trismus

PURPOSEnTo identify imaged regions in which dose is associated with radiation-induced trismus after head and neck cancer radiation therapy (HNRT) using a novel image-based data mining (IBDM) framework.nnnMETHODS AND MATERIALSnA cohort of 86 HNRT patients were analyzed for region identification. Trismus was characterized as a continuous variable by the maximum incisor-to-incisor opening distance (MID) at 6 months after radiation therapy. Patient anatomies and dose distributions were spatially normalized to a common frame of reference using deformable image registration. IBDM was used to identify clusters of voxels associated with MID (P ≤ .05 based on permutation testing). The result was externally tested on a cohort of 35 patients with head and neck cancer. Internally, we also performed a dose-volume histogram-based analysis by comparing the magnitude of the correlation between MID and the mean dose for the IBDM-identified cluster in comparison with 5 delineated masticatory structures.nnnRESULTSnA single cluster was identified with the IBDM approach (P < .01), partially overlapping with the ipsilateral masseter. The dose-volume histogram-based analysis confirmed that the IBDM cluster had the strongest association with MID, followed by the ipsilateral masseter and the ipsilateral medial pterygoid (Spearmans rank correlation coefficients: Rs = -0.36, -0.35, -0.32; P = .001, .001, .002, respectively). External validation confirmed an association between mean dose to the IBDM cluster and MID (Rs = -0.45; P = .007).nnnCONCLUSIONSnIBDM bypasses the common assumption that dose patterns within structures are unimportant. Our novel IBDM approach for continuous outcome variables successfully identified a cluster of voxels that are highly associated with trismus, overlapping partially with the ipsilateral masseter. Tests on an external validation cohort showed an even stronger correlation with trismus. These results support use of the region in HNRT treatment planning to potentially reduce trismus.

Prospective evaluation of relationships between radiotherapy dose to masticatory apparatus and trismus

Abstract Aims: This feasibility study aimed to identify relationships between radiation doses to the masticatory apparatus as a combined block or as individual subunits with changes in trismus following radiotherapy. Material and methods: Twenty patients from a single center were recruited prospectively as part of a randomized trial comparing proactive exercises in the management of trismus. Patients with stage III/IV oral cavity or oropharyngeal squamous cell cancers received intensity-modulated radiotherapy with concurrent systemic therapy. All patients had trismus prior to radiotherapy. Maximal inter-incisor distance (MID) was measured pre- and 6 months from the start of radiotherapy. Bilateral muscles of mastication: medial and lateral pterygoids (MP and LP), masseters (M), temporalis (T), temporomandibular joint (TMJ) were contoured on CT images. The block comprised all muscles excluding the TMJ below the orbital floor. Mean dose, equivalent uniform dose (EUD) and V35–V60u2009Gy were compared with change in MID. Results: In six patients, the MID deteriorated at 6 months from the start of radiotherapy compared with 14 whose MID improved. No significant association was observed between age, gender, smoking, alcohol status, exercise compliance, cisplatin, tumor site, stage, V35–V60u2009Gy or EUD with change in MID. A clinical outlier was excluded. Without the outlier (nu2009=u200919), a significant association was seen between mean dose and change in MID at 6 months for the ipsilateral block (pu2009=u2009.01), LP (pu2009=u2009.04) and M (pu2009<u2009.01). All patients where trismus deteriorated at 6 months received mean doses >40u2009Gy to the block. Conclusion: Higher mean radiation doses to the ipsilateral block, LP and M were significantly associated with deterioration in trismus. Limiting dose to these structures to ≤40u2009Gy for tumors not invading the masticatory muscles may improve treatment-related sequelae. The ipsilateral block, LP and M should be studied further as possible alternative avoidance structures in radiotherapy treatment planning.

EP-1898: A workflow for automatic QA of contour propagation for adaptive radiotherapy

Material and Methods: Free-breathing, sagittal, dynamic multi-slice T2-weighted MRI series of the liver were acquired on a 1.5T scanner (Siemens Avanto) in five healthy volunteers with a balanced steady state free precession sequence (TrueFISP, 20 slices, 20 dynamics, 1.28x1.28x5 mm resolution, 150 msec per slice). Slices were then retrospectively sorted in 4D volumes according to an imagebased method. A volumetric axial T1-weighted acquisition was also performed at breath-hold during inhalation (VIBE, 60 slices, 1.25x1.25x4mm resolution). The proposed method involved applying the motion field derived from the T2weighted 4D MRI dataset to the T1-weighted breath-hold acquisition. Specifically, a rigid registration of the breathhold acquisition was performed onto the T2-weighted series at the corresponding inhale phase. Then, we performed a deformable registration between each respiratory phase and the inhale phase of the T2-weighted 4D scan. The derived motion fields for all respiratory phases were then used to warp the T1-weighted breath hold acquisition (i.e. deriving the virtual T1-weighted 4D MRI).