Impact of target volume segmentation accuracy and variability on treatment planning for 4D-CT-based non-small cell lung cancer radiotherapy

A novel approach for estimating lung tumor motion based on dynamic features in 4D-CT

Due to the high expenses involved, 4D-CT data for certain patients may only include five respiratory phases (0%, 20%, 40%, 60%, and 80%). This limitation can affect the subsequent planning of radiotherapy due to the absence of lung tumor information for the remaining five respiratory phases (10%, 30%, 50%, 70%, and 90%). This study aims to develop an interpolation method that can automatically derive tumor boundary contours for the five omitted phases using the available 5-phase 4D-CT data. The dynamic mode decomposition (DMD) method is a data-driven and model-free technique that can extract dynamic information from high-dimensional data. It enables the reconstruction of long-term dynamic patterns using only a limited number of time snapshots. The quasi-periodic motion of a deformable lung tumor caused by respiratory motion makes it suitable for treatment using DMD. The direct application of the DMD method to analyze the respiratory motion of the tumor is impractical because the tumor is three-dimensional and spans multiple CT slices. To predict the respiratory movement of lung tumors, a method called uniform angular interval (UAI) sampling was developed to generate snapshot vectors of equal length, which are suitable for DMD analysis. The effectiveness of this approach was confirmed by applying the UAI-DMD method to the 4D-CT data of ten patients with lung cancer. The results indicate that the UAI-DMD method effectively approximates the lung tumor's deformable boundary surface and nonlinear motion trajectories. The estimated tumor centroid is within 2 mm of the manually delineated centroid, a smaller margin of error compared to the traditional BSpline interpolation method, which has a margin of 3 mm. This methodology has the potential to be extended to reconstruct the 20-phase respiratory movement of a lung tumor based on dynamic features from 10-phase 4D-CT data, thereby enabling more accurate estimation of the planned target volume (PTV).

Improving normal tissue sparing using scripting in endometrial cancer radiation therapy planning

The aim of this study was to improve the protection of organs at risk (OARs), decrease the total planning time and maintain sufficient target doses using scripting endometrial cancer external beam radiation therapy (EBRT) planning. Computed tomography (CT) data of 14 endometrial cancer patients were included in this study. Manual and automatic planning with scripting were performed for each CT. Scripts were created in the RayStation™ (RaySearch Laboratories AB, Stockholm, Sweden) planning system using a Python code. In scripting, seven additional contours were automatically created to reduce the OAR doses. The scripted and manual plans were compared to each other in terms of planning time, dose-volume histogram (DVH) parameters, and total monitor unit (MU) values. While the mean total planning time for manual planning was 368 ± 8 s, it was only 55 ± 2 s for the automatic planning with scripting (p < 0.001). The mean doses of OARs decreased with automatic planning (p < 0.001). In addition, the maximum doses (D2% and D1%) for bilateral femoral heads and the rectum were significantly reduced. It was observed that the total MU value increased from 1146 ± 126 (manual planning) to 1369 ± 95 (scripted planning). It is concluded that scripted planning has significant time and dosimetric advantages over manual planning for endometrial cancer EBRT planning.

Probabilistic evaluation of plan quality for time-dependent anatomical deformations in head and neck cancer patients

PURPOSE

In addition to patient set-up uncertainties, anatomical deformations, e.g., weight loss, lead to time-dependent differences between the planned and delivered dose in a radiotherapy course that currently cannot easily be predicted. The aim of this study was to create time-varying prediction models to describe both the average and residual anatomical deformations.

METHODS

Weekly population-based principal component analysis models were generated from on-treatment cone-beam CT scans (CBCTs) of 30 head and neck cancer patients, with additional data of 35 patients used as a validation cohort. We simulated treatment courses accounting for a) anatomical deformations, b) set-up uncertainties and c) a combination of both. The dosimetric effects of the simulated deformations were compared to a direct dose accumulation based on deformable registration of the CBCT data.

RESULTS

Set-up uncertainties were seen to have a larger effect on the organ at risk (OAR) doses than anatomical deformations for all OARs except the larynx and the primary CTV. Distributions from simulation results were in good agreement with those of the accumulated dose.

CONCLUSIONS

We present a novel method of modelling time-varying organ deformations in head and neck cancer. The effect on the OAR doses from these deformations are smaller than the effect of set-up uncertainties for most OARs. These models can, for instance, be used to predict which patients could benefit from adaptive radiotherapy, prior to commencing treatment.

Erring Characteristics of Deformable Image Registration-Based Auto-Propagation for Internal Target Volume in Radiotherapy of Locally Advanced Non-Small Cell Lung Cancer

Purpose

Respiratory motion of locally advanced non-small cell lung cancer (LA-NSCLC) adds to the challenge of targeting the disease with radiotherapy (RT). One technique used frequently to alleviate this challenge is an internal gross tumor volume (IGTV) generated from manual contours on a single respiratory phase of the 4DCT via the aid of deformable image registration (DIR)-based auto-propagation. Through assessing the accuracy of DIR-based auto-propagation for generating IGTVs, this study aimed to identify erring characteristics associated with the process to enhance RT targeting in LA-NSCLC.

Methods

4DCTs of 19 patients with LA-NSCLC were acquired using retrospective gating with 10 respiratory phases (RPs). Ground-truth IGTVs (GT-IGTVs) were obtained through manual segmentation and union of gross tumor volumes (GTVs) in all 10 phases. IGTV auto-propagation was carried out using two distinct DIR algorithms for the manually contoured GTV from each of the 10 phases, resulting in 10 separate IGTVs for each patient per each algorithm. Differences between the auto-propagated IGTVs (AP-IGTVs) and their corresponding GT-IGTVs were assessed using Dice coefficient (DICE), maximum symmetric surface distance (MSSD), average symmetric surface distance (ASSD), and percent volume difference (PVD) and further examined in relation to anatomical tumor location, RP, and deformation index (DI) that measures the degree of deformation during auto-propagation. Furthermore, dosimetric implications due to the analyzed differences between the AP-IGTVs and GT-IGTVs were assessed.

Results

Findings were largely consistent between the two algorithms: DICE, MSSD, ASSD, and PVD showed no significant differences between the 10 RPs used for propagation (Kruskal–Wallis test, ps &gt; 0.90); MSSD and ASSD differed significantly by tumor location in the central–peripheral and superior–inferior dimensions (ps &lt; 0.0001) while only in the central–peripheral dimension for PVD (p &lt; 0.001); DICE, MSSD, and ASSD significantly correlated with the DI (Spearman’s rank correlation test, ps &lt; 0.0001). Dosimetric assessment demonstrated that 79% of the radiotherapy plans created by targeting planning target volumes (PTVs) derived from the AP-IGTVs failed prescription constraints for their corresponding ground-truth PTVs.

Conclusion

In LA-NSCLC, errors in DIR-based IGTV propagation present to varying degrees and manifest dependences on DI and anatomical tumor location, indicating the need for personalized consideration in designing RT internal target volume.

Geometric and Dosimetric Evaluation of the Automatic Delineation of Organs at Risk (OARs) in Non-Small-Cell Lung Cancer Radiotherapy Based on a Modified DenseNet Deep Learning Network

Purpose

To introduce an end-to-end automatic segmentation model for organs at risk (OARs) in thoracic CT images based on modified DenseNet, and reduce the workload of radiation oncologists.

Materials and Methods

The computed tomography (CT) images of 36 lung cancer patients were included in this study, of which 27 patients’ images were randomly selected as the training set, 9 patients’ as the testing set. The validation set was generated by cross validation and 6 patients’ images were randomly selected from the training set during each epoch as the validation set. The autosegmentation task of the left and right lungs, spinal cord, heart, trachea and esophagus was implemented, and the whole training time was approximately 5 hours. Geometric evaluation metrics including the Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95) and average surface distance (ASD), were used to assess the autosegmentation performance of OARs based on the proposed model and were compared with those based on U-Net as benchmarks. Then, two sets of treatment plans were optimized based on the manually contoured targets and OARs (Plan1), as well as the manually contours targets and the automatically contoured OARs (Plan2). Dosimetric parameters, including Dmax, Dmean and Vx, of OARs were obtained and compared.

Results

The DSC, HD95 and ASD of the proposed model were better than those of U-Net. The differences in the DSC of the spinal cord and esophagus, differences in the HD95 of the spinal cord, heart, trachea and esophagus, as well as differences in the ASD of the spinal cord were statistically significant between the two models (P<0.05). The differences in the dose-volume parameters of the two sets of plans were not statistically significant (P>0.05). Moreover, compared with manual segmentation, autosegmentation significantly reduced the contouring time by nearly 40.7% (P<0.05).

Conclusions

The bilateral lungs, spinal cord, heart and trachea could be accurately delineated using the proposed model in this study; however, the automatic segmentation effect of the esophagus must still be further improved. The concept of feature map reuse provides a new idea for automatic medical image segmentation.

ProCaLung - Peer review in stage III, mediastinal node-positive, non-small-cell lung cancer: How to benchmark clinical practice of nodal target volume definition and delineation in Belgium☆

BACKGROUND AND PURPOSE

The Quality Assurance project for stage III non-small cell lung cancer radiotherapy ProCaLung performed a multicentric two-step exercise evaluating mediastinal nodal Target Volume Definition and Delineation (TVD) variability and the opportunity for standardization. The TVD variability before and after providing detailed guidelines and the value of qualitative contour reviewing before applying quantitative measures were investigated.

MATERIALS AND METHODS

The case of a patient with stage III NSCLC and involved mediastinal lymph nodes was used as a basis for this study. Twenty-two radiation oncologists from nineteen centers in Belgium and Luxembourg participated in at least one of two phases of the project (before and after introduction of ProCaLung contouring guidelines). The resulting thirty-three mediastinal nodal GTV and CTV contours were then evaluated using a qualitative-before-quantitative (QBQ) approach. First, a qualitative analysis was performed, evaluating adherence to most recent guidelines. From this, a list of observed deviations was created and these were used to evaluate contour conformity. The second analysis was quantitative, using overlap and surface distance measures to compare contours within qualitative groups and between phases. A 'most robust' reference volume for these analyses was created using the STAPLE-algorithm and an averaging method.

RESULTS

Five GTV and seven CTV qualitative groups were identified. Second step contours were more often in higher-conformity groups (p = 0.012 for GTV and p = 0.024 for CTV). Median Residual Mean Square Distances improved from 2.34 mm to 1.36 mm for GTV (p = 0.01) and from 4.53 mm to 1.58 mm for CTV (p < 0.0001). Median Dice coefficients increased from 0.81 to 0.84 for GTV (p = 0.07) and from 0.82 to 0.89 for CTV (p ≤ 0.001). Using HC-contours only to generate references translated in more robust quantitative evaluations.

CONCLUSION

Variability of mediastinal nodal TVD was reduced after providing the ProCaLung consensus guidelines. A qualitative review was essential for providing meaningful quantitative measures.

Lung tumor segmentation in 4D CT images using motion convolutional neural networks

PURPOSE

Manual delineation on all breathing phases of lung cancer 4D CT image datasets can be challenging, exhaustive, and prone to subjective errors because of both the large number of images in the datasets and variations in the spatial location of tumors secondary to respiratory motion. The purpose of this work is to present a new deep learning-based framework for fast and accurate segmentation of lung tumors on 4D CT image sets.

METHODS

The proposed DL framework leverages motion region convolutional neural network (R-CNN). Through integration of global and local motion estimation network architectures, the network can learn both major and minor changes caused by tumor motion. Our network design first extracts tumor motion information by feeding 4D CT images with consecutive phases into an integrated backbone network architecture, locating volume-of-interest (VOIs) via a regional proposal network and removing irrelevant information via a regional convolutional neural network. Extracted motion information is then advanced into the subsequent global and local motion head network architecture to predict corresponding deformation vector fields (DVFs) and further adjust tumor VOIs. Binary masks of tumors are then segmented within adjusted VOIs via a mask head. A self-attention strategy is incorporated in the mask head network to remove any noisy features that might impact segmentation performance. We performed two sets of experiments. In the first experiment, a five-fold cross-validation on 20 4D CT datasets, each consisting of 10 breathing phases (i.e., 200 3D image volumes in total). The network performance was also evaluated on an additional unseen 200 3D images volumes from 20 hold-out 4D CT datasets. In the second experiment, we trained another model with 40 patients' 4D CT datasets from experiment 1 and evaluated on additional unseen nine patients' 4D CT datasets. The Dice similarity coefficient (DSC), center of mass distance (CMD), 95th percentile Hausdorff distance (HD95 ), mean surface distance (MSD), and volume difference (VD) between the manual and segmented tumor contour were computed to evaluate tumor detection and segmentation accuracy. The performance of our method was quantitatively evaluated against four different methods (VoxelMorph, U-Net, network without global and local networks, and network without attention gate strategy) across all evaluation metrics through a paired t-test.

RESULTS

The proposed fully automated DL method yielded good overall agreement with the ground truth for contoured tumor volume and segmentation accuracy. Our model yielded significantly better values of evaluation metrics (p < 0.05) than all four competing methods in both experiments. On hold-out datasets of experiment 1 and 2, our method yielded DSC of 0.86 and 0.90 compared to 0.82 and 0.87, 0.75 and 0.83, 081 and 0.89, and 0.81 and 0.89 yielded by VoxelMorph, U-Net, network without global and local networks, and networks without attention gate strategy. Tumor VD between ground truth and our method was the smallest with the value of 0.50 compared to 0.99, 1.01, 0.92, and 0.93 for between ground truth and VoxelMorph, U-Net, network without global and local networks, and networks without attention gate strategy, respectively.

CONCLUSIONS

Our proposed DL framework of tumor segmentation on lung cancer 4D CT datasets demonstrates a significant promise for fully automated delineation. The promising results of this work provide impetus for its integration into the 4D CT treatment planning workflow to improve the accuracy and efficiency of lung radiotherapy.

4D CT analysis of organs at risk (OARs) in stereotactic radiotherapy

Internal organs at risk volumes (IRV) represent the propagation of organs at risk (OARs) in 4DCT. Sixty consecutive patients that underwent 4DCT for thoracic stereotactic radiotherapy were analyzed and IRVs for heart, trachea, esophagus, bronchial tree, great vessels, and spinal cord were calculated. IRVs were then tested for the respect of dose constraints. IRVs were significantly bigger than standard OARs (p-value <0.001 for all the IRVs). IRVs that did not respect the dose constraints were, respectively, 7/60 (11.7%) for Heart IRV, 6/60 (10%) for Esophagus IRV, 11/60 (18.3%) for Trachea IRV, 16/60 (26.6%) for Bronchial Tree and 0/60 (0%) for great vessel and spinal cord IRV. In the subset of central targets, the percentage of plans that can be unacceptable taking into consideration OARs motion reaches 42%. The correlation of IRVs with clinical parameters and toxicity deserves future investigations in prospective trials.

Quantifying the dosimetric impact of organ-at-risk delineation variability in head and neck radiation therapy in the context of patient setup uncertainty

The purpose of this study was to quantify the potential dosimetric impact of delineation variability (DV) in head and neck radiation therapy (RT) when inherent patient setup variability (SV) is also considered. The impact of DV was assessed by generating plans with multiple structure sets, cross-evaluating them, including SV, across sets, and determining P _PQM: the probability of achieving organ-specific plan quality metrics (PQM). DV was incorporated by: (1) using multiple organ at risk (OAR) structure sets delineated by independent manual observers; and (2) randomly perturbing manually generated OARs to generate alternatives with varying levels of uncertainty (low, medium, and high DV). For each structure set, independent VMAT plans were auto-generated to meet clinical PQMs. Each plan was cross-evaluated using OARs from multiple structure sets with simulated SV including per-fraction random (σ _s) and per-treatment-course systematic (Σ_s) setup errors. The dosimetric impact of DV was assessed by examining P _PQM with and without SV/DV. Clinically significant differences were defined by those that exceeded differences caused by a  +2% output variation. Without including SV, simulated DV at the medium level reduced P _PQM by an average of 5.5% for all OARs with D _max PQMs. This reduction decreased to 2.8% for SV  =  2 mm and 2.4% for SV  =  4 mm (the average P _PQM reduction due to 2% output errors was 2.7%). For OARs with D _mean PQMs, the average P _PQM reduction was 0.9% for SV  =  0 and  ⩽0.1% for SV  ⩾  2 mm. The effect of DV was larger for OARs that directly abutted a target volume than for those that did not. These trends were also observed with real DV from multi-observer delineations. The dosimetric impact of DV appeared to decrease when random and systematic SV was considered. Sensitivity to DV was affected by OAR objective type (i.e. D _mean versus D _max objectives) as well as distance from the target volume.

A review of automatic lung tumour segmentation in the era of 4DCT

Aim

To review the literature on auto-contouring methods of lung tumour volumes on four-dimensional computed tomography (4DCT).

Background

Manual delineation of lung tumour on 4DCT has been the gold standard in clinical practice. However, it is resource intensive due to the high volume of data which results in longer contouring duration and uncertainties in defining target. Auto-contouring may present as an attractive alternative by decreasing manual inputs required, thus improving the contouring process. This review aims to assess the accuracy, variability and contouring duration of automatic contouring compared with manual contouring in lung cancer on 4DCT datasets.

Materials and methods

A search and review of literature were conducted to identify studies regarding lung tumour contouring on 4DCT. Manual and auto-contours were assessed and compared based on accuracy, variability and contouring duration.

Results

Thirteen studies were included in this review and their results were compared. Accuracy of auto-contours was found to be comparable to manual contours. Auto-contouring resulted in lesser inter-observer variation when compared to manual contouring, however there was no significant reduction in intra-observer variability. Additionally, contouring duration was reduced with auto-contouring although long computation time could present as a bottleneck.

Conclusion

Auto-contouring is reliable and efficient, producing accurate contours with better consistency compared to manual contours. However, manual inputs would still be required both before and after auto-propagation.

Evaluation of automatic contour propagation in T2-weighted 4DMRI for normal-tissue motion assessment using internal organ-at-risk volume (IRV)

Purpose

The purpose of this study was to evaluate the quality of automatically propagated contours of organs at risk (OARs) based on respiratory‐correlated navigator‐triggered four‐dimensional magnetic resonance imaging (RC‐4DMRI) for calculation of internal organ‐at‐risk volume (IRV) to account for intra‐fractional OAR motion.

Methods and Materials

T2‐weighted RC‐4DMRI images were of 10 volunteers acquired and reconstructed using an internal navigator‐echo surrogate and concurrent external bellows under an IRB‐approved protocol. Four major OARs (lungs, heart, liver, and stomach) were delineated in the 10‐phase 4DMRI. Two manual‐contour sets were delineated by two clinical personnel and two automatic‐contour sets were propagated using free‐form deformable image registration. The OAR volume variation within the 10‐phase cycle was assessed and the IRV was calculated as the union of all OAR contours. The OAR contour similarity between the navigator‐triggered and bellows‐rebinned 4DMRI was compared. A total of 2400 contours were compared to the most probable ground truth with a 95% confidence level (S95) in similarity, sensitivity, and specificity using the simultaneous truth and performance level estimation (STAPLE) algorithm.

Results

Visual inspection of automatically propagated contours finds that approximately 5–10% require manual correction. The similarity, sensitivity, and specificity between manual and automatic contours are indistinguishable (P > 0.05). The Jaccard similarity indexes are 0.92 ± 0.02 (lungs), 0.89 ± 0.03 (heart), 0.92 ± 0.02 (liver), and 0.83 ± 0.04 (stomach). Volume variations within the breathing cycle are small for the heart (2.6 ± 1.5%), liver (1.2 ± 0.6%), and stomach (2.6 ± 0.8%), whereas the IRV is much larger than the OAR volume by: 20.3 ± 8.6% (heart), 24.0 ± 8.6% (liver), and 47.6 ± 20.2% (stomach). The Jaccard index is higher in navigator‐triggered than bellows‐rebinned 4DMRI by 4% (P < 0.05), due to the higher image quality of navigator‐based 4DMRI.

Conclusion

Automatic and manual OAR contours from Navigator‐triggered 4DMRI are not statistically distinguishable. The navigator‐triggered 4DMRI image provides higher contour quality than bellows‐rebinned 4DMRI. The IRVs are 20–50% larger than OAR volumes and should be considered in dose estimation.

ACTIVE MEAN FIELDS FOR PROBABILISTIC IMAGE SEGMENTATION: CONNECTIONS WITH CHAN-VESE AND RUDIN-OSHER-FATEMI MODELS

Segmentation is a fundamental task for extracting semantically meaningful regions from an image. The goal of segmentation algorithms is to accurately assign object labels to each image location. However, image-noise, shortcomings of algorithms, and image ambiguities cause uncertainty in label assignment. Estimating the uncertainty in label assignment is important in multiple application domains, such as segmenting tumors from medical images for radiation treatment planning. One way to estimate these uncertainties is through the computation of posteriors of Bayesian models, which is computationally prohibitive for many practical applications. On the other hand, most computationally efficient methods fail to estimate label uncertainty. We therefore propose in this paper the Active Mean Fields (AMF) approach, a technique based on Bayesian modeling that uses a mean-field approximation to efficiently compute a segmentation and its corresponding uncertainty. Based on a variational formulation, the resulting convex model combines any label-likelihood measure with a prior on the length of the segmentation boundary. A specific implementation of that model is the Chan-Vese segmentation model (CV), in which the binary segmentation task is defined by a Gaussian likelihood and a prior regularizing the length of the segmentation boundary. Furthermore, the Euler-Lagrange equations derived from the AMF model are equivalent to those of the popular Rudin-Osher-Fatemi (ROF) model for image denoising. Solutions to the AMF model can thus be implemented by directly utilizing highly-efficient ROF solvers on log-likelihood ratio fields. We qualitatively assess the approach on synthetic data as well as on real natural and medical images. For a quantitative evaluation, we apply our approach to the icgbench dataset.