Intra- and Inter-operator variability in manual tumor segmentation: Impact on radionuclide therapy dosimetry (original) (raw)
Related papers
EJNMMI Physics, 2022
The aim was to quantify inter-and intra-observer variability in manually delineated hepatocellular carcinoma (HCC) lesion contours and the resulting impact on radioembolization (RE) dosimetry. Methods: Ten patients with HCC lesions treated with Y-90 RE and imaged with post-therapy Y-90 PET/CT were selected for retrospective analysis. Three radiologists contoured 20 lesions manually on baseline multiphase contrast-enhanced MRIs, and two of the radiologists re-contoured at two additional sessions. Contours were transferred to co-registered PET/CT-based Y-90 dose maps. Volume-dependent recovery coefficients were applied for partial volume correction (PVC) when reporting mean absorbed dose. To understand how uncertainty varies with tumor size, we fit power models regressing relative uncertainty in volume and in mean absorbed dose on contour volume. Finally, we determined effects of segmentation uncertainty on tumor control probability (TCP), as calculated using logistic models developed in a previous RE study. Results: The average lesion volume ranged from 1.8 to 194.5 mL, and the mean absorbed dose ranged from 23.4 to 1629.0 Gy. The mean inter-observer Dice coefficient for lesion contours was significantly less than the mean intra-observer Dice coefficient (0.79 vs. 0.85, p < 0.001). Uncertainty in segmented volume, as measured by the Coefficient of Variation (CV), ranged from 4.2 to 34.7% with an average of 17.2%. The CV in mean absorbed dose had an average value of 5.4% (range 1.2-13.1%) without PVC while it was 15.1% (range 1.5-55.2%) with PVC. Using the fitted models for uncertainty as a function of volume on our prior data, the mean change in TCP due to segmentation uncertainty alone was estimated as 16.2% (maximum 48.5%). Conclusions: Though we find relatively high inter-and intra-observer reliability overall, uncertainty in tumor contouring propagates into non-negligible uncertainty in dose metrics and outcome prediction for individual cases that should be considered in dosimetry-guided treatment.
Radiotherapy and Oncology, 2016
" and ''intraobserver" to identify articles which evaluated interobserver variability in target or organ-at-risk (OAR) volume delineation. Studies had to fulfil the following criteria to be selected for this review: Table 1 Summary of studies of inter-observer variability in volume delineation. Site Volume analysed No of studies No of studies with type of datasets Median no of scans (range) Median no of observers (range) Metrics used No of studies evaluating dosimetry No of studies using statistical analysis References
Australasian Physics & Engineering Sciences in Medicine, 2001
Uncertainty in the precise quantity of radiation dose delivered to tumours in external beam radiotherapy is present due to many factors, and can result in either spatially uniform (Gaussian) or spatially non-uniform dose errors. These dose errors are incorporated into the calculation of tumour control probability (TCP) and produce a distribution of possible TCP values over a population. We also study the effect of inter-patient cell sensitivity heterogeneity on the population distribution of patient TCPs. This study aims to investigate the relative importance of these three uncertainties (spatially uniform dose uncertainty, spatially non-uniform dose uncertainty, and inter-patient cell sensitivity heterogeneity) on the delivered dose and TCP distribution following a typical course of fractionated external beam radiotherapy. The dose distributions used for patient treatments are modelled in one dimension. Geometric positioning uncertainties during and before treatment are considered as shifts of a pre-calculated dose distribution. Following the simulation of a population of patients, distributions of dose across the patient population are used to calculate mean treatment dose, standard deviation in mean treatment dose, mean TCP, standard deviation in TCP, and TCP mode. These parameters are calculated with each of the three uncertainties included separately. The calculations show that the dose errors in the tumour volume are dominated by the spatially uniform component of dose uncertainty. This could be related to machine specific parameters, such as linear accelerator calibration. TCP calculation is affected dramatically by inter-patient variation in the cell sensitivity and to a lesser extent by the spatially uniform dose errors. The positioning errors with the 1.5 em margins used cause dose uncertainty outside the tumour volume and have a small effect on mean treatment dose (in the tumour volume) and tumour control.
EJNMMI Physics
Background: The purpose was to validate 90 Y PET gradient-based tumor segmentation in phantoms and to evaluate the impact of the segmentation method on reported tumor absorbed dose (AD) and biological effective dose (BED) in 90 Y microsphere radioembolization (RE) patients. A semi-automated gradient-based method was applied to phantoms and patient tumors on the 90 Y PET with the initial bounding volume for gradient detection determined from a registered diagnostic CT or MR; this PET-based segmentation (PS) was compared with radiologist-defined morphologic segmentation (MS) on CT or MRI. AD and BED volume histogram metrics (D90, D70, mean) were calculated using both segmentations and concordance/correlations were investigated. Spatial concordance was assessed using Dice similarity coefficient (DSC) and mean distance to agreement (MDA). PS was repeated to assess intra-observer variability. Results: In phantoms, PS demonstrated high accuracy in lesion volumes (within 15%), AD metrics (within 11%), high spatial concordance relative to morphologic segmentation (DSC > 0.86 and MDA < 1.5 mm), and low intra-observer variability (DSC > 0.99, MDA < 0. 2 mm, AD/BED metrics within 2%). For patients (58 lesions), spatial concordance between PS and MS was degraded compared to in-phantom (average DSC = 0.54, average MDA = 4.8 mm); the average mean tumor AD was 226 ± 153 and 197 ± 138 Gy, respectively for PS and MS. For patient AD metrics, the best Pearson correlation (r) and concordance correlation coefficient (ccc) between segmentation methods was found for mean AD (r = 0.94, ccc = 0.92), but worsened as the metric approached the minimum dose (for D90, r = 0.77, ccc = 0.69); BED metrics exhibited a similar trend. Patient PS showed low intra-observer variability (average DSC = 0.81, average MDA = 2.2 mm, average AD/BED metrics within 3.0%). Conclusions: 90 Y PET gradient-based segmentation led to accurate/robust results in phantoms, and showed high concordance with MS for reporting mean tumor AD/BED in patients. However, tumor coverage metrics such as D90 exhibited worse concordance between segmentation methods, highlighting the need to standardize segmentation methods when reporting AD/BED metrics from post-therapy 90 Y PET. Estimated differences in reported AD/BED metrics due to segmentation method will be useful for interpreting RE dosimetry results in the literature including tumor response data.
2003
To evaluate the impact of interobserver variability in the contouring of gross tumor volumes (GTVs) and clinical target volumes (CTVs) on the global geometric accuracy in radiation therapy. Material and Methods: In a review of the currently available literature, the magnitude of interobserver variability is analyzed, causes and consequences are discussed. Uncertainties due to inconsistencies in contouring are related to other sources of geometric errors, particularly patient positioning and organ motion. Results: Interobserver variability is a major -for some tumor locations probably the largest -factor contributing to geometric inaccuracy. Causes are multifactorial and include image-and observer-related factors, such as the subjective interpretation of image information. Conclusion: Consequences to reduce interobserver variability are proposed, among others the selection of adequate imaging modalities, intensified radiologic training, and the use of telecommunication tools.
Delineation uncertainties of tumour volumes on MRI of head and neck cancer patients
Clinical and Translational Radiation Oncology
Background: During the last decade, radiotherapy using MR Linac has gone from research to clinical implementation for different cancer locations. For head and neck cancer (HNC), target delineation based only on MR images is not yet standard, and the utilisation of MRI instead of PET/CT in radiotherapy planning is not well established. We aimed to analyse the inter-observer variation (IOV) in delineating GTV (gross tumour volume) on MR images only for patients with HNC. Material/methods: 32 HNC patients from two independent departments were included. Four clinical oncologists from Denmark and four radiation oncologists from Australia had independently contoured primary tumour GTVs (GTV-T) and nodal GTVs (GTV-N) on T2-weighted MR images obtained at the time of treatment planning. Observers were provided with sets of images, delineation guidelines and patient synopsis. Simultaneous truth and performance level estimation (STAPLE) reference volumes were generated for each structure using all observer contours. The IOV was assessed using the DICE Similarity Coefficient (DSC) and mean absolute surface distance (MASD). Results: 32 GTV-Ts and 68 GTV-Ns were contoured per observer. The median MASD for GTV-Ts and GTV-Ns across all patients was 0.17 cm (range 0.08-0.39 cm) and 0.07 cm (range 0.04-0.33 cm), respectively. Median DSC relative to a STAPLE volume for GTV-Ts and GTV-Ns across all patients were 0.73 and 0.76, respectively. A significant correlation was seen between median DSCs and median volumes of GTV-Ts (Spearman correlation coefficient 0.76, p < 0.001) and of GTV-Ns (Spearman correlation coefficient 0.55, p < 0.001). Conclusion: Contouring GTVs in patients with HNC on MRI showed that the median IOV for GTV-T and GTV-N was below 2 mm, based on observes from two separate radiation departments. However, there are still specific regions in tumours that are difficult to resolve as either malignant tissue or oedema that potentially could be improved by further training in MR-only delineation.
Margins for geometric uncertainty around organs at risk in radiotherapy
Radiotherapy and Oncology, 2002
Background and purpose: ICRU Report 62 suggests drawing margins around organs at risk (ORs) to produce planning organ at risk volumes (PRVs) to account for geometric uncertainty in the radiotherapy treatment process. This paper proposes an algorithm for drawing such margins, and compares the recommended margin widths with examples from clinical practice and discusses the limitations of the approach. Method: The use of the PRV defined in this way is that, despite the geometric uncertainties, the dose calculated within the PRV by the treatment planning system can be used to represent the dose in the OR with a certain confidence level. A suitable level is where, in the majority of cases (90%), the dose-volume histogram of the PRV will not under-represent the high-dose components in the OR. In order to provide guidelines on how to do this in clinical practice, this paper distinguishes types of OR in terms of the tolerance doses relative to the prescription dose and suggests appropriate margins for serial-structure and parallel-structure ORs. Results: In some instances of large and parallel ORs, the clinician may judge that the complication risk in omitting a margin is acceptable. Otherwise, for all types of OR, systematic, treatment preparation uncertainties may be accommodated by an OR ! PRV margin width of 1.3S. Here, S is the standard deviation of the combined systematic (treatment preparation) uncertainties. In the case of serial ORs or small, parallel ORs, the effects of blurring caused by daily treatment execution errors (set-up and organ motion) should be taken into account. Near a region of high dose, blurring tends to shift the isodoses away from the unblurred edge as shown on the treatment planning system by an amount that may be represented by 0.5s. This margin may be used either to increase or to decrease the margin already calculated for systematic uncertainties, depending upon the size of the tolerance dose relative to the detailed planned dose distribution. Where the detailed distribution is unknown before the OR is delineated, then the overall margin for serial or small parallel ORs should be 1.3S 1 0.5s. Examples are given where the application of this algorithm leads to margin widths around ORs similar to those in use clinically. Conclusions: Using PRVs is appropriate both for forward and inverse planning. Dose-volume histograms of PRVs for serial-and parallelstructure ORs require careful interpretation. Nevertheless, use of the proposed algorithms for drawing margins around both serial and parallel ORs can alert the dosimetrist/radiation oncologist to the possibility of high-dose complications in individual treatment plans, which might be missed if no such margins were drawn.
Radiotherapy and Oncology, 1993
The enormous developments in radiation technology open new horizons for improvements in local tumour control. However, the evolution from conventional external beam radiotherapy planning to conformal therapy might be hampered by the potential risk of over-reliance on the physician's capability of estimating the tumour extent from imaging modalities. The variability between 12 volunteering physicians in the delineation of tumour and target volume on the lateral orthogonal localisation radiograph from CT was assessed for 5 brain tumours. The estimated tumour and target sizes varied, respectively with a factor of 1.3-2.6 and with a factor of 1.3-2.1. The anatomical location of the volumes showed maximum variations from 11 to 27 mm in the crania-caudal direction and from 14 to 21 mm in the fronto-occipital direction. For the 5 test cases, the tumour area on which all radiation oncologists agreed, represented only 25-73X of the corresponding mean tumour area. Although the introduction of computed tomography in radiation treatment planning was proved to be a major step forwards for treatment planning in many tumour sites, the results of the present study on brain tumours demonstrate that the subjective interpretation of the tumour extent based on CT images might be one of the largest factors contributing to the overall uncertainty in radiation treatment planning. Moreover, this study endorses the need for uncertainty analysis of the medical decision-making process. It may be that the process of making uncertainties explicit can contribute to the improvement of our present concept of radiation treatment planning. However, more data are urgently needed to estimate the uncertainty in target volume delineation in different tumour sites and with different imaging modalities such as MRI. One of the most exciting and contributing factors in accurate delineation of tumours for radiation treatment in the future might be in the field of computer-aided image correlation.
2023
Objective: Advances in hepatic radioembolization are based on a selective approach with radical intent and the use of multicompartment dosimetric analysis. The objective of this study is to assess the utility of voxel-based dosimetry in the quantification of actual absorbed doses in radiation segmentectomy procedures and to establish cutoff values predictive of response. Methods: Ambispective study in hepatocarcinoma patients treated with radiation segmentectomy. Calculated dosimetric parameters were mean tumor-absorbed dose, maximum tumor AD, minimal tumor AD in 30, 50, and 70% of tumor volume and mean AD in non-tumor liver. The actual absorbed dose (aAD) was calculated on the Y-90-PET/ CT image using 3D voxel-based dosimetry software. To assess radiological response, localized mRECIST criteria were used. The objective response rate (ORR) was defined as CR or PR. Results: Twenty-four HCC patients, BCLC 0 (5), A (17) and B (2) were included. The mean yttrium-90 administered activity was 1.38 GBq in a mean angiosome volume of 206.9 cc and tumor volume 56.01 cc. The mean theoretical AD was 306.3 Gy and aAD 352 Gy. A very low concordance was observed between both parameters (rho_c 0.027). ORR at 3 and 6 m was 84.21% and 92.31%, respectively. Statistically significant relationship was observed between the maximum tumor-absorbed dose and complete radiological response at 3 m (p 0.022). Conclusion: A segmental approach with radical intention leads to response rates greater than 90%, being the tumor maximum absorbed dose the dosimetric parameter that best predicts radiological response in voxel-based dosimetry.