Optimization in treatment planning of high dose-rate brachytherapy - Review and analysis of mathematical models

Penalty weight tuning in high dose rate brachytherapy using multi-objective Bayesian optimization

Objective.Treatment plan optimization in high dose rate brachytherapy often requires manual fine-tuning of penalty weights for each objective, which can be time-consuming and dependent on the planner's experience. To automate this process, this study used a multi-criteria approach called multi-objective Bayesian optimization with q-noisy expected hypervolume improvement as its acquisition function (MOBO-qNEHVI).Approach.The treatment plans of 13 prostate cancer patients were retrospectively imported to a research treatment planning system, RapidBrachyMTPS, where fast mixed integer optimization (FMIO) performs dwell time optimization given a set of penalty weights to deliver 15 Gy to the target volume. MOBO-qNEHVI was used to find patient-specific Pareto optimal penalty weight vectors that yield clinically acceptable dose volume histogram metrics. The relationship between the number of MOBO-qNEHVI iterations and the number of clinically acceptable plans per patient (acceptance rate) was investigated. The performance time was obtained for various parameter configurations.Main results.MOBO-qNEHVI found clinically acceptable treatment plans for all patients. With increasing the number of MOBO-qNEHVI iterations, the acceptance rate grew logarithmically while the performance time grew exponentially. Fixing the penalty weight of the tumour volume to maximum value, adding the target dose as a parameter, initiating MOBO-qNEHVI with 25 parallel sampling of FMIO, and running 6 MOBO-qNEHVI iterations found solutions that delivered 15 Gy to the hottest 95% of the clinical target volume while respecting the dose constraints to the organs at risk. The average acceptance rate for each patient was 89.74% ± 8.11%, and performance time was 66.6 ± 12.6 s. The initiation took 22.47 ± 7.57 s, and each iteration took 7.35 ± 2.45 s to find one Pareto solution.Significance.MOBO-qNEHVI combined with FMIO can automatically explore the trade-offs between treatment plan objectives in a patient specific manner within a minute. This approach can reduce the dependency of plan quality on planner's experience and reduce dose to the organs at risk.

Treatment Planning Strategies for Interstitial Ultrasound Ablation of Prostate Cancer

PURPOSE

To develop patient-specific 3D models using Finite-Difference Time-Domain (FDTD) simulations and pre-treatment planning tools for the selective thermal ablation of prostate cancer with interstitial ultrasound. This involves the integration with a FDA 510(k) cleared catheter-based ultrasound interstitial applicators and delivery system.

METHODS

A 3D generalized "prostate" model was developed to generate temperature and thermal dose profiles for different applicator operating parameters and anticipated perfusion ranges. A priori planning, based upon these pre-calculated lethal thermal dose and iso-temperature clouds, was devised for iterative device selection and positioning. Full 3D patient-specific anatomic modeling of actual placement of single or multiple applicators to conformally ablate target regions can be applied, with optional integrated pilot-point temperature-based feedback control and urethral/rectum cooling. These numerical models were verified against previously reported ex-vivo experimental results obtained in soft tissues.

RESULTS

For generic prostate tissue, 360 treatment schemes were simulated based on the number of transducers (1-4), applied power (8-20 W/cm2), heating time (5, 7.5, 10 min), and blood perfusion (0, 2.5, 5 kg/m3/s) using forward treatment modelling. Selectable ablation zones ranged from 0.8-3.0 cm and 0.8-5.3 cm in radial and axial directions, respectively. 3D patient-specific thermal treatment modeling for 12 Cases of T2/T3 prostate disease demonstrate applicability of workflow and technique for focal, quadrant and hemi-gland ablation. A temperature threshold (e.g., Tthres = 52 °C) at the treatment margin, emulating placement of invasive temperature sensing, can be applied for pilot-point feedback control to improve conformality of thermal ablation. Also, binary power control (e.g., Treg = 45 °C) can be applied which will regulate the applied power level to maintain the surrounding temperature to a safe limit or maximum threshold until the set heating time.

CONCLUSIONS

Prostate-specific simulations of interstitial ultrasound applicators were used to generate a library of thermal-dose distributions to visually optimize and set applicator positioning and directivity during a priori treatment planning pre-procedure. Anatomic 3D forward treatment planning in patient-specific models, along with optional temperature-based feedback control, demonstrated single and multi-applicator implant strategies to effectively ablate focal disease while affording protection of normal tissues.

The use of deep learning in interventional radiotherapy (brachytherapy): A review with a focus on open source and open data

Deep learning advanced to one of the most important technologies in almost all medical fields. Especially in areas, related to medical imaging it plays a big role. However, in interventional radiotherapy (brachytherapy) deep learning is still in an early phase. In this review, first, we investigated and scrutinised the role of deep learning in all processes of interventional radiotherapy and directly related fields. Additionally, we summarised the most recent developments. For better understanding, we provide explanations of key terms and approaches to solving common deep learning problems. To reproduce results of deep learning algorithms both source code and training data must be available. Therefore, a second focus of this work is on the analysis of the availability of open source, open data and open models. In our analysis, we were able to show that deep learning plays already a major role in some areas of interventional radiotherapy, but is still hardly present in others. Nevertheless, its impact is increasing with the years, partly self-propelled but also influenced by closely related fields. Open source, data and models are growing in number but are still scarce and unevenly distributed among different research groups. The reluctance in publishing code, data and models limits reproducibility and restricts evaluation to mono-institutional datasets. The conclusion of our analysis is that deep learning can positively change the workflow of interventional radiotherapy but there is still room for improvements when it comes to reproducible results and standardised evaluation methods.

Plan quality score to evaluate the dwell time deviation restricted inverse planning by simulated annealing and graphically optimized treatment plans for template based interstitial brachytherapy

PURPOSE

To evaluate the impact of dwell time deviation constraint (DTDC) on the quality of IPSA-optimized treatment plans in comparison with graphical plans using plan quality scores (PQS).

MATERIAL AND METHODS

Seventy optimized plans (graphical & IPSA with different DTDC values) of ten cervical cancer patients were generated. Various DVH parameters like D₉₀, V₁₀₀, V₁₅₀, V₂₀₀, V₃₀₀ were compared to evaluate the impact of DTDC on target coverage and high dose regions inside target for different plans. Similarly, for the OAR dose, values of D_2cc were compared. Various planning parameters like CI, COIN, DHI, DNR, ODI, EI and gain factor (GF) for different OARs were calculated. Based on these indices a plan quality score (PQS) was formulated and calculated. PQS values were used to see the impact of DTDC on plan quality of IPSA in comparison with dosimetric quality of graphical plan.

RESULTS

We have found that target coverage is similar for IPSA and graphically optimized treatment plans. However, dose homogeneity was improved in IPSA compared to graphical optimization whereas conformality was better in graphically optimized plans. OAR dose was less in IPSA plans. High-dose regions inside the target were also reduced in IPSA comparatively. However, IPSA plans optimized with various values of DTDC did not necessarily reduce high-dose regions beyond 0.6. Plan quality scores (PQS) were 6.31, 6.31, 6.34, and 6.17 for the graphically optimized plan, IPSA with DTDC values of 0.0, 0.4, and 1.0 respectively.

CONCLUSION

We found that IPSA is dosimetrically advantageous over graphical optimization. IPSA with a DTDC value of 0.4 improved overall plan quality. However, DTDC value beyond 0.6 produces dosimetrically sub-optimal plans hence the use of DTDC should be very selective and limited.

Validation of automated post-adjustments of HDR prostate brachytherapy treatment plans by quantitative measures and oncologist observer study

PURPOSE

The aim was to evaluate a postprocessing optimization algorithm's ability to improve the spatial properties of a clinical treatment plan while preserving the target coverage and the dose to the organs at risk. The goal was to obtain a more homogenous treatment plan, minimizing the need for manual adjustments after inverse treatment planning.

MATERIALS AND METHODS

The study included 25 previously treated prostate cancer patients. The treatment plans were evaluated on dose-volume histogram parameters established clinical and quantitative measures of the high dose volumes. The volumes of the four largest hot spots were compared and complemented with a human observer study with visual grading by eight oncologists. Statistical analysis was done using ordinal logistic regression. Weighted kappa and Fleiss' kappa were used to evaluate intra- and interobserver reliability.

RESULTS

The quantitative analysis showed that there was no change in planning target volume (PTV) coverage and dose to the rectum. There were significant improvements for the adjusted treatment plan in: V150% and V200% for PTV, dose to urethra, conformal index, and dose nonhomogeneity ratio. The three largest hot spots for the adjusted treatment plan were significantly smaller compared to the clinical treatment plan. The observers preferred the adjusted treatment plan in 132 cases and the clinical in 83 cases. The observers preferred the adjusted treatment plan on homogeneity and organs at risk but preferred the clinical plan on PTV coverage.

CONCLUSIONS

Quantitative analysis showed that the postadjustment optimization tool could improve the spatial properties of the treatment plans while maintaining the target coverage.

Technical note: Evaluation of a spatial optimization model for prostate high dose-rate brachytherapy in a clinical treatment planning system

BACKGROUND

Spatial properties of a dose distribution, such as volumes of contiguous hot spots, are of clinical importance in treatment planning for high dose-rate brachytherapy (HDR BT). We have in an earlier study developed an optimization model that reduces the prevalence of contiguous hot spots by modifying a tentative treatment plan.

PURPOSE

The aim of this study is to incorporate the correction of hot spots in a standard inverse planning workflow and to validate the integrated model in a clinical treatment planning system. The spatial function is included in the objective function for the inverse planning, as opposed to in the previous study where it was applied as a separate post-processing step. Our aim is to demonstrate that fine-adjustments of dose distributions, which are often performed manually in today's clinical practice, can be automated.

METHODS

A spatial optimization function was introduced in the treatment planning system RayStation (RaySearch Laboratories AB, Stockholm, Sweden) via a research interface. A series of 10 consecutive prostate patients treated with HDR BT was retrospectively replanned with and without the spatial function.

RESULTS

Optimization with the spatial function decreased the volume of the largest contiguous hot spot by on average 31%, compared to if the function was not included. The volume receiving at least 200% of the prescription dose decreased by on average 11%. Target coverage, measured as the fractions of the clinical target volume (CTV) and the planning target volume (PTV) receiving at least the prescription dose, was virtually unchanged (less than a percent change for both metrics). Organs-at-risk received comparable or slightly decreased doses if the spatial function was included in the optimization model.

CONCLUSIONS

Optimization of spatial properties such as the volume of contiguous hot spots can be integrated in a standard inverse planning workflow for brachytherapy, and need not be conducted as a separate post-processing step.

Dosimetric impact of a robust optimization approach to mitigate effects from rotational uncertainty in prostate intensity-modulated brachytherapy

BACKGROUND

Intensity-modulated brachytherapy (IMBT) is an emerging technology for cancer treatment, in which radiation sources are shielded to shape the dose distribution. The rotatable shields provide an additional degree of freedom, but also introduce an additional, directional, type of uncertainty, compared to conventional high-dose-rate brachytherapy (HDR BT).

PURPOSE

We propose and evaluate a robust optimization approach to mitigate the effects of rotational uncertainty in the shields with respect to planning criteria.

METHODS

A previously suggested prototype for platinum-shielded prostate ¹⁶⁹ Yb-based dynamic IMBT is considered. We study a retrospective patient data set (anatomical contours and catheter placement) from two clinics, consisting of six patients that had previously undergone conventional ¹⁹² Ir HDR BT treatment. The Monte Carlo-based treatment planning software RapidBrachyMCTPS is used for dose calculations. In our computational experiments, we investigate systematic rotational shield errors of ±10° and ±20°, and the same systematic error is applied to all dwell positions in each scenario. This gives us three scenarios, one nominal and two with errors. The robust optimization approach finds a compromise between the average and worst-case scenario outcomes.

RESULTS

We compare dose plans obtained from standard models and their robust counterparts. With dwell times obtained from a linear penalty model (LPM), for 10° errors, the dose to urethra ( D 0.1 c c $D_{0.1cc}$ ) and rectum ( D 0.1 c c $D_{0.1cc}$ and D 1 c c $D_{1cc}$ ) increase with up to 5% and 7%, respectively, in the worst-case scenario, while with the robust counterpart, the corresponding increases were 3% and 3%. For all patients and all evaluated criteria, the worst-case scenario outcome with the robust approach had lower deviation compared to the standard model, without compromising target coverage. We also evaluated shield errors up to 20° and while the deviations increased to a large extent with the standard models, the robust models were capable of handling even such large errors.

CONCLUSIONS

We conclude that robust optimization can be used to mitigate the effects from rotational uncertainty and to ensure the treatment plan quality of IMBT.

A Novel Workflow with a Customizable 3D Printed Vaginal Template and a Direction Modulated Brachytherapy (DMBT) Tandem Applicator for Adaptive Interstitial Brachytherapy of the Cervix

A novel clinical workflow utilizing a direction modulated brachytherapy (DMBT) tandem applicator in combination with a patient-specific, 3D printed vaginal needle-track template for an advanced image-guided adaptive interstitial brachytherapy of the cervix. The proposed workflow has three main steps: (1) pre-treatment MRI, (2) an initial optimization of the needle positions based on the DMBT tandem positioning and patient anatomy, and a subsequent inverse optimization using the combined DMBT tandem and needles, and (3) rapid 3D printing. We retrospectively re-planned five patient cases for two scenarios; one plan with the DMBT tandem (T) and ovoids (O) with the original needle (ND) positions (DMBT + O + ND) and another with the DMBT T&O and spatially reoptimized needles (OptN) positions (DMBT + O + OptN). All retrospectively reoptimized plans have been compared to the original plan (OP) as well. The accuracy of 3D printing was verified through the image registration between the planning CT and the CT of the 3D-printed template. The average difference in D_2cc for the bladder, rectum, and sigmoid between the OPs and DMBT + O + OptNs were -8.03 ± 4.04%, -18.67 ± 5.07%, and -26.53 ± 4.85%, respectively. In addition, these average differences between the DMBT + O + ND and DMBT + O + OptNs were -2.55 ± 1.87%, -10.70 ± 3.45%, and -22.03 ± 6.01%, respectively. The benefits could be significant for the patients in terms of target coverage and normal tissue sparing and increase the optimality over free-hand needle positioning.

Comparative Analysis of 60Co and 192Ir Sources in High Dose Rate Brachytherapy for Cervical Cancer

High-dose-rate (HDR) brachytherapy (BT) is an essential treatment for cervical cancer, one of the most prevalent gynecological malignant tumors. In HDR BT, high radiation doses can be delivered to the tumor target with the minimum radiation doses to organs at risk. Despite the wide use of the small HDR 192Ir source, as the technique has improved, the HDR 60Co source, which has the same miniaturized geometry, has also been produced and put into clinical practice. Compared with 192Ir (74 days), 60Co has a longer half-life (5.3 years), which gives it a great economic advantage for developing nations. The aim of the study was to compare 60Co and 192Ir sources for HDR BT in terms of both dosimetry and clinical treatment. The results of reports published on the use of HDR BT for cervical cancer over the past few years as well as our own research show that this treatment is safe and it is feasible to use 60Co as an alternative source.

Dose Optimization for Single-Channel Vaginal Cylinder High-Dose-Rate Brachytherapy: A Double Prescription Method for Patients With Endometrial Adenocarcinoma

Purpose

This study aimed to explore the relationship between applicator surface dose and 5 mm-depth dose and to optimize both locations simultaneously for three most used cylinder sizes (2.5, 3.0, and 3.5 cm in diameter) in treating patients with endometrial adenocarcinoma.

Materials and methods

A total of 216 plans were created for each dose level and applicator size. For each dose level, four plans were created with single or double prescription doses. For plans with double prescription doses, the dose constraints were applied to all those points on the surface and 5 mm depth and optimize the two sites simultaneously. 

Results

A dose table between surface and 5 mm depth and its fifth order polynomial mapping functions were established for each applicator size, so any prescribed dose at one site can find the prescription dose on the other site in optimization on both locations. For plans with a 5 mm-depth prescription, the maximum dose on the surface can be reduced from 145% to 133% if the surface prescription dose is also used; for plans with surface dose prescription, the minimum dose and mean dose can be improved by 2% if 5 mm-depth dose prescription is also used in optimization.

Conclusion

Dose table and their mapping functions between surface prescription dose and their corresponding 5 mm-depth doses were created. A new optimization method that uses two prescription doses on both surface and 5 mm-depth sites was proposed to reduce the hot dose on the surface and improve the cold dose at 5 mm depth.

Incorporating Treatment Time into Butterfly Optimization to Reduce Total Treatment Time for Vaginal Cylinder Brachytherapy

Purpose

For patient comfort and safety, irradiation times should be kept at a minimum while maintaining high treatment quality. In this study of high dose rate (HDR) therapy with a vaginal cylinder, we used the butterfly optimization algorithm (BOA) to simultaneously optimize individual dwell times for precise dose conformity and for the reduction of total dwell time.

Material and methods

BOA is a population-based, meta-heuristic algorithm that averts local minima by conducting intensive local and global searching based on switching probability. We constructed an objective function (a stimulus intensity function) that consisted of two components. The first one was the root-mean-squared dose error (RMSE) defined as the square root of the sum of squared differences between the prescribed and delivered dose at the constraint points. The second component was weighted total treatment time. Eight previously treated cases were retrospectively reviewed by re-optimizing the clinical treatment plans with BOA. 

Results

Compared to the eight original plans generated with the commercial adaptive volume optimization algorithm (AVOA), the BOA-optimized plans reduced treatment times by 5.4% to 8.9%, corresponding to a time-saving of 13.1 to 47.7 seconds with the activities on the treatment day and saving from 29.3 to 64.6 seconds if treated with an activity of 5 CI. Dose deviations from the prescription were smaller than in the original plans.

Conclusion 

Dose optimizations based on the BOA algorithm yield closer dose conformity in vaginal HDR treatment than AVOA. Incorporating total treatment time into the optimization algorithm reduces the delivery time while having only a small effect on dose conformity.

Deep reinforcement learning for treatment planning in high-dose-rate cervical brachytherapy

PURPOSE

High-dose-rate (HDR) brachytherapy (BT) is an effective cancer treatment method in which the radiation source is placed within the body. Treatment planning is a critical component for a successful outcome. Almost all currently proposed treatment planning methods are built on stochastic heuristic algorithms, which limits the generation of higher quality plans. This study proposed a novel treatment planning method to adjust dwell times in a human-like fashion to improve the quality of the plan.

METHODS

We built an intelligent treatment planner network (ITPN) based on deep reinforcement learning (DRL). The network architecture of ITPN is Dueling Double-Deep Q Network. The state is the dwell time of each dwell position and the action is which dwell time to adjust and how to adjust it. A hybrid equivalent uniform dose objective function was established and assigned corresponding rewards according to its changes. Experience replay was performed with the epsilon greedy algorithm and SumTree data structure.

RESULTS

In the evaluation of ITPN using 20 patient cases, D₉₀, D₁₀₀ and V₁₀₀ showed no significant difference compared with inverse planning simulated annealing (IPSA) optimization. However, D_2cc of bladder, rectum and sigmoid, V₁₅₀ and V₂₀₀ were significant reduced, and homogeneity index and conformity index were significantly increased.

CONCLUSION

The proposed ITPN was able to generate higher quality plans based on the learned dwell time adjustment policy than IPSA. This is the first artificial intelligence system that can directly determine the dwell times of HDR BT, which demonstrated the potential feasibility of solving optimization problems via DRL.

Emerging technologies in brachytherapy

Brachytherapy is a mature treatment modality. The literature is abundant in terms of review articles and comprehensive books on the latest established as well as evolving clinical practices. The intent of this article is to part ways and look beyond the current state-of-the-art and review emerging technologies that are noteworthy and perhaps may drive the future innovations in the field. There are plenty of candidate topics that deserve a deeper look, of course, but with practical limits in this communicative platform, we explore four topics that perhaps is worthwhile to review in detail at this time. First, intensity modulated brachytherapy (IMBT) is reviewed. The IMBT takes advantage ofanisotropicradiation profile generated through intelligent high-density shielding designs incorporated onto sources and applicators such to achieve high quality plans. Second, emerging applications of 3D printing (i.e. additive manufacturing) in brachytherapy are reviewed. With the advent of 3D printing, interest in this technology in brachytherapy has been immense and translation swift due to their potential to tailor applicators and treatments customizable to each individual patient. This is followed by, in third, innovations in treatment planning concerning catheter placement and dwell times where new modelling approaches, solution algorithms, and technological advances are reviewed. And, fourth and lastly, applications of a new machine learning technique, called deep learning, which has the potential to improve and automate all aspects of brachytherapy workflow, are reviewed. We do not expect that all ideas and innovations reviewed in this article will ultimately reach clinic but, nonetheless, this review provides a decent glimpse of what is to come. It would be exciting to monitor as IMBT, 3D printing, novel optimization algorithms, and deep learning technologies evolve over time and translate into pilot testing and sensibly phased clinical trials, and ultimately make a difference for cancer patients. Today's fancy is tomorrow's reality. The future is bright for brachytherapy.