Evaluation of inter-observer variability of bladder boundary delineation on cone-beam CT

Artificial intelligence-generated targets and inter-observer variation in online adaptive radiotherapy of bladder cancer

Background and purpose

Daily target re-delineation in online adaptive radiotherapy (oART) introduces uncertainty. The aim of this study was to evaluate artificial intelligence (AI) generated contours and inter-observer target variation among radiotherapy technicians in cone-beam CT (CBCT) guided oART of bladder cancer.

Materials and methods

For each of 10 consecutive patients treated with oART for bladder cancer, one CBCT was randomly selected and retrospectively included. The bladder (CTV-T) was AI-segmented (CTV-TAI). Seven radiotherapy technicians independently reviewed and edited CTV-TAI, generating CTV-TADP. Contours were benchmarked against a ground truth contour (CTV-TGT) delineated blindly from scratch. CTV-TADP and CTV-TAI were compared to CTV-TGT using volume, dice similarity coefficient, and bidirectional local distance. Dose coverage (D99%>95 %) of CTV-TGT was evaluated for treatment plans optimized for CTV-TAI and CTV-TADP with clinical margins. Inter-observer variation among CTV-TADP was assessed using coefficient of variation and generalized conformity index.

Results

CTV-TGT ranged from 48.7 cm3 to 211.6 cm3. The median [range] volume difference was 4.5 [-17.8, 42.4] cm3 for CTV-TADP and -15.5 [-54.2, 4.3] cm3 for CTV-TAI, compared to CTV-TGT. Corresponding dice similarity coefficients were 0.87 [0.71, 0.95] and 0.84 [0.64, 0.95]. CTV-TGT was adequately covered in 68/70 plans optimized on CTV-TADP and in 6/10 plans optimized on CTV-TAI with clinical margins. The median [range] coefficient of variation was 0.08 [0.05, 0.11] and generalized conformity index was 0.78 [0.71, 0.88] among CTV-TADP.

Conclusions

Target re-delineation in CBCT-guided oART of bladder cancer demonstrated non-isotropic inter-observer variation. Manual adjustment of AI-generated contours was necessary to cover ground truth targets.

Online adaptive radiotherapy for bladder cancer using a simultaneous integrated boost and fiducial markers

PURPOSE

The aim was to assess the feasibility of online adaptive radiotherapy (oART) for bladder cancer using a focal boost by focusing on the quality of the online treatment plan and automatic target delineation, duration of the workflow and performance in the presence of fiducial markers for tumor bed localization.

METHODS

Fifteen patients with muscle invasive bladder cancer received daily oART with Cone Beam CT (CBCT), artificial intelligence (AI)-assisted automatic delineation of the daily anatomy and online plan reoptimization. The bladder and pelvic lymph nodes received a total dose of 40 Gy in 20 fractions, the tumor received an additional simultaneously integrated boost (SIB) of 15 Gy. The dose distribution of the reference plan was calculated for the daily anatomy, i.e. the scheduled plan. Simultaneously, a reoptimization of the plan was performed i.e. the adaptive plan. The target coverage and V95% outside the target were evaluated for both plans. The need for manual adjustments of the GTV delineation, the duration of the workflow and the influence of fiducial markers were assessed.

RESULTS

All 300 adaptive plans met the requirement of the CTV-coverage V95%≥98% for both the boost (55 Gy) and elective volume (40 Gy). For the scheduled plans the CTV-coverage was 53.5% and 98.5%, respectively. Significantly less tissue outside the targets received 55 Gy in case of the adaptive plans as compared to the scheduled plans. Manual corrections of the GTV were performed in 67% of the sessions. In 96% of these corrections the GTV was enlarged and resulted in a median improvement of 1% for the target coverage. The median on-couch time was 22 min. A third of the session time consisted of reoptimization of the treatment plan. Fiducial markers were visible on the CBCTs and aided the tumor localization.

CONCLUSIONS

AI-driven CBCT-guided oART aided by fiducial markers is feasible for bladder cancer radiotherapy treatment including a SIB. The quality of the adaptive plans met the clinical requirements and fiducial markers were visible enabling consistent daily tumor localization. Improved automatic delineation to lower the need for manual corrections and faster reoptimization would result in shorter session time.

Recommendations for planning and delivery of radical radiotherapy for localized urothelial carcinoma of the bladder

PURPOSE

Curative radio-chemotherapy is recognized as a standard treatment option for muscle-invasive bladder cancer (MIBC). Nevertheless, the technical aspects for MIBC radiotherapy are heterogeneous with a lack of practical recommendations.

METHODS AND MATERIALS

In 2018, a workshop identified the need for two cooperative groups to develop consistent, evidence-based guidelines for irradiation technique in the delivery of curative radiotherapy. Two radiation oncologists performed a review of the literature addressing several topics relative to radical bladder radiotherapy: planning computed tomography acquisition, target volume delineation, radiation schedules (total dose and fractionation) and dose delivery (including radiotherapy techniques, image-guided radiotherapy (IGRT) and adaptive treatment modalities). Searches for original and review articles in the PubMed and Google Scholar databases were conducted from January 1990 until March 2020. During a meeting conducted in October 2020, results on 32 topics were presented and discussed with a working group involving 15 radiation oncologists, 3 urologists and one medical oncologist. We applied the American Urological Association guideline development's method to define a consensus strategy.

RESULTS

A consensus was obtained for all 34 except 4 items. The group did not obtain an agreement on CT enhancement added value for planning, PTV margins definition for empty bladder and full bladder protocols, and for pelvic lymph-nodes irradiation. High quality evidence was shown in 6 items; 8 items were considered as low quality of evidence.

CONCLUSION

The current recommendations propose a homogenized modality of treatment both for routine clinical practice and for future clinical trials, following the best evidence to date, analyzed with a robust methodology. The XXX group formulates practical guidelines for the implementation of innovative techniques such as adaptive radiotherapy.

Image-guided Adaptive Radiotherapy for Bladder Cancer

Technological advancement has facilitated patient-specific radiotherapy in bladder cancer. This has been made possible by developments in image-guided radiotherapy (IGRT). Particularly transformative has been the integration of volumetric imaging into the workflow. The ability to visualise the bladder target using cone beam computed tomography and magnetic resonance imaging initially assisted with determining the magnitude of inter- and intra-fraction target change. It has led to greater confidence in ascertaining true anatomy at each fraction. The increased certainty of dose delivered to the bladder has permitted the safe reduction of planning target volume margins. IGRT has therefore improved target coverage with a reduction in integral dose to the surrounding tissue. Use of IGRT to feed back into plan and dose delivery optimisation according to the anatomy of the day has enabled adaptive radiotherapy bladder solutions. Here we undertake a review of the stepwise developments underpinning IGRT and adaptive radiotherapy strategies for external beam bladder cancer radiotherapy. We present the evidence in accordance with the framework for systematic clinical evaluation of technical innovations in radiation oncology (R-IDEAL).

MR-Guided Adaptive Radiotherapy for Bladder Cancer

Radiotherapy has an important role in the curative and palliative treatment settings for bladder cancer. As a target for radiotherapy the bladder presents a number of technical challenges. These include poor tumor visualization and the variability in bladder size and position both between and during treatment delivery. Evidence favors the use of magnetic resonance imaging (MRI) as an important means of tumor visualization and local staging. The availability of hybrid systems incorporating both MRI scanning capabilities with the linear accelerator (MR-Linac) offers opportunity for in-room and real-time MRI scanning with ability of plan adaption at each fraction while the patient is on the treatment couch. This has a number of potential advantages for bladder cancer patients. In this article, we examine the technical challenges of bladder radiotherapy and explore how magnetic resonance (MR) guided radiotherapy (MRgRT) could be leveraged with the aim of improving bladder cancer patient outcomes. However, before routine clinical implementation robust evidence base to establish whether MRgRT translates into improved patient outcomes should be ascertained.

Role of deformable image registration for delivered dose accumulation of adaptive external beam radiation therapy and brachytherapy in cervical cancer

Purpose

Deformable image registration (DIR) can be used to accumulate the absorbed dose distribution of daily image-guided adaptive external beam radiation treatment (EBRT) and brachytherapy (BT). Since dose-volume parameter addition assumes a uniform delivered EBRT dose around the planned BT boost, the added value of DIR over direct addition was investigated for dose accumulation in bladder and rectum.

Material and methods

For 10 patients (EBRT 46/46.2 Gy_EQD2, EBRT + BT: D₉₀ 85-90 Gy_EQD2, in equivalent dose in 2 Gy fractions), the actually delivered dose from adaptive volumetric-modulated arc therapy (VMAT)/intensity-modulated radiotherapy (IMRT) EBRT was calculated using the daily anatomy from the cone-beam computed tomography (CBCT) scans acquired prior to irradiation. The CBCT of the first EBRT fraction and the BT planning MRI were registered using DIR. The cumulative dose to the 2 cm³ with the highest dose (D_2cm3) from EBRT and BT to the bladder and rectum was calculated and compared to direct addition assuming a uniform EBRT dose (UD).

Results

Differences (DIR-UD) in the total EBRT + BT dose ranged between –0.2-3.9 Gy_EQD2 (bladder) and –1.0-3.7 Gy_EQD2 (rectum). The total EBRT + BT dose calculated with DIR was at most 104% of the dose calculated with the UD method.

Conclusions

Differences between UD and DIR were small (< 3.9 Gy_EQD2). The dose delivered with adaptive VMAT/IMRT EBRT to bladder and rectum near the planned BT boost can be considered uniform for the evaluation of bladder/rectum D_2cm3.

LEFMIS: locally-oriented evaluation framework for medical image segmentation algorithms

This article proposes a novel framework for the locally-oriented evaluation of segmentation algorithms (LEFMIS). The presented approach is robust and takes into account local inter/intra-observer variability and the anisotropy of medical images. What is more, the framework makes it possible to distinguish types of error locally. These features are crucial in the context of cancer image data. The proposed framework is based on use of the signed anisotropic Euclidean distance transform and the distance projection. It can be used easily in many different applications with or without additional expert outlines (both inter- and intra-observer variability). The performance of the proposed framework is depicted using both artificial and kidney cancer CT data with experts' manual outlines. In the article, in the case of artificial data, it is presented that the manual outlines dispersion is symmetric in relation to the truth border. The effectiveness of the selected segmentation algorithm was analysed in the context of kidney cancer using computed tomography data. For the calculated local inter-observer variability, 80.11% of the surface points generated by the kidney segmentation algorithm are within one expert outline standard deviation and 97.96% are within five. An error distribution shift in the direction of type I error equivalent was also observed. Finally, the significance of the local estimation of error type differences is presented. The article shows the greater usefulness and flexibility of the proposed framework in comparison to the state-of-the-art methods. The exemplary usage of the LEFMIS with or without inter-/intra-observer variability is also presented.

Image guided adaptive external beam radiation therapy for cervix cancer: Evaluation of a clinically implemented plan-of-the-day technique

BACKGROUND

Radiotherapy for cervix cancer is challenging in patients exhibiting large daily changes in the pelvic anatomy, therefore adaptive treatments (ART) have been proposed. The aim of this study was the clinical implementation and subsequent evaluation of plan-of-the-day (POTD)-ART for cervix cancer in supine positioning. The described workflow was based on standard commercial equipment and current quality assurance (QA) methods.

MATERIALS AND METHODS

A POTD strategy, which employs a VMAT plan library consisting of an empty bladder plan, a full bladder plan and a motion robust backup plan, was developed. Daily adaption was guided by cone beam computed tomography (CBCT) imaging after which the best plan from the library was selected. Sixteen patients were recruited in a clinical study on ART, for nine POTD was applied due to their large organ motion derived from two computed tomography (CT) scans with variable bladder filling. All patients were treated to 45Gy in 25 fractions. Plan selection frequencies over the treatment course were analyzed. Daily doses in the rectum, bladder and cervix-uterus target (CTV-T) were derived and compared to a simulated non-adapted treatment (non-ART), which employed the robust plan for each fraction. Additionally, the adaption consistency was determined by repeating the plan selection procedure one month after treatment by a group of experts. ART-specific QA methods are presented.

RESULTS

225 ART fractions with CBCTs were analyzed. The empty bladder plan was delivered in 49% of the fractions in the first treatment week and this number increased to 78% in the fifth week. The daily coverage of the CTV-T was equivalent between ART and the non-ART simulation, while the daily total irradiated volume V42.75Gy (95% of prescription dose) was reduced by a median of 87cm³. The median delivered V42.75Gy was 1782cm³. Daily delivered doses (V42.75Gy, V40Gy, V30G) to the organs at risk were statistically significantly reduced by ART, with a median difference in daily V42.75Gy in rectum and bladder of 3.2% and 1.1%, respectively. The daily bladder V42.75Gy and V40Gy were decreased by more than 10 percent points in 30% and 24% of all fractions, respectively, through ART. The agreement between delivered plans and retrospective expert-group plan selections was 84%.

CONCLUSION

A POTD-ART technique for cervix cancer was successfully and safely implemented in the clinic and evaluated. Improved normal tissue sparing compared to a simulated non-ART treatment could be demonstrated. Future developments should focus on commercial automated software solutions to allow for a more widespread adoption and to keep the increased workload manageable.

A comparative assessment of prostate positioning guided by three-dimensional ultrasound and cone beam CT

Background

The accuracy of the Elekta Clarity™ three-dimensional ultrasound system (3DUS) was assessed for prostate positioning and compared to seed- and bone-based positioning in kilo-voltage cone-beam computed tomography (CBCT) during a definitive radiotherapy.

Methods

The prostate positioning of 6 patients, with fiducial markers implanted into the prostate, was controlled by 3DUS and CBCT. In total, 78 ultrasound scans were performed trans-abdominally and compared to bone-matches and seed-matches in CBCT scans. Setup errors detected by the different modalities were compared. Systematic and random errors were analysed, and optimal setup margins were calculated.

Results

The discrepancy between 3DUS and seed-match in CBCT was −0.2 ± 2.7 mm laterally, −1.9 ± 2.3 mm longitudinally and 0.0 ± 3.0 mm vertically and significant only in longitudinal direction. Using seed-match as reference, systematic errors of 3DUS were 1.3 mm laterally, 0.8 mm longitudinally and 1.4 mm vertically, and random errors were 2.5 mm laterally, 2.3 mm longitudinally, and 2.7 mm vertically. No significant difference could be detected for 3DUS in comparison to bone-match in CBCT.

Conclusions

3DUS is feasible for image guidance for patients with prostate cancer and appears comparable to CBCT based image guidance in the retrospective study. While 3DUS offers some distinct advantages such as no need of invasive fiducial implantation and avoidance of extra radiation, its disadvantages include the operator dependence of the technique and dependence on sufficient bladder filling. Further study of 3DUS for image guidance in a large patient cohort is warranted.

Kinetic Models for Predicting Cervical Cancer Response to Radiation Therapy on Individual Basis Using Tumor Regression Measured In Vivo With Volumetric Imaging

This article describes a macroscopic mathematical modeling approach to capture the interplay between solid tumor evolution and cell damage during radiotherapy. Volume regression profiles of 15 patients with uterine cervical cancer were reconstructed from serial cone-beam computed tomography data sets, acquired for image-guided radiotherapy, and used for model parameter learning by means of a genetic-based optimization. Patients, diagnosed with either squamous cell carcinoma or adenocarcinoma, underwent different treatment modalities (image-guided radiotherapy and image-guided chemo-radiotherapy). The mean volume at the beginning of radiotherapy and the end of radiotherapy was on average 23.7 cm(3) (range: 12.7-44.4 cm(3)) and 8.6 cm(3) (range: 3.6-17.1 cm(3)), respectively. Two different tumor dynamics were taken into account in the model: the viable (active) and the necrotic cancer cells. However, according to the results of a preliminary volume regression analysis, we assumed a short dead cell resolving time and the model was simplified to the active tumor volume. Model learning was performed both on the complete patient cohort (cohort-based model learning) and on each single patient (patient-specific model learning). The fitting results (mean error: ∼ 16% and ∼ 6% for the cohort-based model and patient-specific model, respectively) highlighted the model ability to quantitatively reproduce tumor regression. Volume prediction errors of about 18% on average were obtained using cohort-based model computed on all but 1 patient at a time (leave-one-out technique). Finally, a sensitivity analysis was performed and the data uncertainty effects evaluated by simulating an average volume perturbation of about 1.5 cm(3) obtaining an error increase within 0.2%. In conclusion, we showed that simple time-continuous models can represent tumor regression curves both on a patient cohort and patient-specific basis; this discloses the opportunity in the future to exploit such models to predict how changes in the treatment schedule (number of fractions, doses, intervals among fractions) might affect the tumor regression on an individual basis.