Radiotherapy Planning Using an Improved Search Strategy in Particle Swarm Optimization

Biological optimization for mediastinal lymphoma radiotherapy - a preliminary study

Purpose: In current radiotherapy (RT) planning and delivery, population-based dose-volume constraints are used to limit the risk of toxicity from incidental irradiation of organs at risks (OARs). However, weighing tradeoffs between target coverage and doses to OARs (or prioritizing different OARs) in a quantitative way for each patient is challenging. We introduce a novel RT planning approach for patients with mediastinal Hodgkin lymphoma (HL) that aims to maximize overall outcome for each patient by optimizing on tumor control and mortality from late effects simultaneously.Material and Methods: We retrospectively analyzed 34 HL patients treated with conformal RT (3DCRT). We used published data to model recurrence and radiation-induced mortality from coronary heart disease and secondary lung and breast cancers. Patient-specific doses to the heart, lung, breast, and target were incorporated in the models as well as age, sex, and cardiac risk factors (CRFs). A preliminary plan of candidate beams was created for each patient in a commercial treatment planning system. From these candidate beams, outcome-optimized (O-OPT) plans for each patient were created with an in-house optimization code that minimized the individual risk of recurrence and mortality from late effects. O-OPT plans were compared to VMAT plans and clinical 3DCRT plans.Results: O-OPT plans generally had the lowest risk, followed by the clinical 3DCRT plans, then the VMAT plans with the highest risk with median (maximum) total risk values of 4.9 (11.1), 5.1 (17.7), and 7.6 (20.3)%, respectively (no CRFs). Compared to clinical 3DCRT plans, O-OPT planning reduced the total risk by at least 1% for 9/34 cases assuming no CRFs and 11/34 cases assuming presence of CRFs.Conclusions: We developed an individualized, outcome-optimized planning technique for HL. Some of the resulting plans were substantially different from clinical plans. The results varied depending on how risk models were defined or prioritized.

Automatic VMAT planning for post-operative prostate cancer cases using particle swarm optimization: A proof of concept study

OBJECTIVE

To investigate the potential of Particle Swarm Optimization (PSO) for fully automatic VMAT radiotherapy (RT) treatment planning.

MATERIAL AND METHODS

In PSO a solution space of planning constraints is searched for the best possible RT plan in an iterative, statistical method, optimizing a population of candidate solutions. To identify the best candidate solution and for final evaluation a plan quality score (PQS), based on dose volume histogram (DVH) parameters, was introduced. Automatic PSO-based RT planning was used for N = 10 postoperative prostate cancer cases, retrospectively taken from our clinical database, with a prescribed dose of EUD = 66 Gy in addition to two constraints for rectum and one for bladder. Resulting PSO-based plans were compared dosimetrically to manually generated VMAT plans.

RESULTS

PSO successfully proposed treatment plans comparable to manually optimized ones in 9/10 cases. The median (range) PTV EUD was 65.4 Gy (64.7-66.0) for manual and 65.3 Gy (62.5-65.5) for PSO plans, respectively. However PSO plans achieved significantly lower doses in rectum D_2% 67.0 Gy (66.5-67.5) vs. 66.1 Gy (64.7-66.5, p = 0.016). All other evaluated parameters (PTV D_98% and D_2%, rectum V_40Gy and V_60Gy, bladder D_2% and V_60Gy) were comparable in both plans. Manual plans had lower PQS compared to PSO plans with -0.82 (-16.43-1.08) vs. 0.91 (-5.98-6.25).

CONCLUSION

PSO allows for fully automatic generation of VMAT plans with plan quality comparable to manually optimized plans. However, before clinical implementation further research is needed concerning further adaptation of PSO-specific parameters and the refinement of the PQS.

Descriptor Selection Improvements for Quantitative Structure-Activity Relationships

Molecular descriptor selection is an essential procedure to improve a predictive quantitative structure–activity relationship (QSAR) model. However, within the QSAR model, there are a number of redundant, noisy and irrelevant descriptors. In this study, we propose a novel descriptor selection framework using self-paced learning (SPL) via sparse logistic regression (LR) with Logsum penalty (SPL-Logsum), which can simultaneously adaptively identify the simple and complex samples and avoid over-fitting. SPL is inspired by the learning process of humans or animals gradually learned from simple and complex samples to train models, and the Logsum penalized LR helps to select a small subset of significant molecular descriptors for improving the QSAR models. Experimental results on some simulations and three public QSAR datasets show that our proposed SPL-Logsum framework outperforms other existing sparse methods regarding the area under the curve, sensitivity, specificity, accuracy, and [Formula: see text]-values.

Inverse radiotherapy planning based on bioeffect modelling for locally advanced left-sided breast cancer

BACKGROUND AND PURPOSE

Treatment planning of radiotherapy (RT) for left-sided breast cancer is a challenging case. Several competing concerns are incorporated at present through protocol-defined dose-volume constraints, e.g. cardiac exposure and target coverage. Such constraints are limited by neglecting patient-specific risk factors (RFs). We propose an alternative RT planning method based solely on bioeffect models to minimize the estimated risks of breast cancer recurrence (BCR) and radiation-induced mortality endpoints considering patient-specific factors.

METHODS AND MATERIALS

Thirty-nine patients with left-sided breast cancer treated with comprehensive post-lumpectomy loco-regional conformal RT were included. An in-house particle swarm optimization (PSO) engine was used to choose fields from a large set of predefined fields and optimize monitor units to minimize the total risk of BCR and mortality caused by radiation-induced ischaemic heart disease (IHD), secondary lung cancer (SLC) and secondary breast cancer (SBC). Risk models included patient age, smoking status and cardiac risk and were developed using published multi-institutional data.

RESULTS

For the clinical plans the normal tissue complication probability, i.e. summed risk of IHD, SLC and SBC, was <3.7% and the risk of BCR was <6.1% for all patients. Median total decrease in mortality or recurrence achieved with individualized PSO plans was 0.4% (range, 0.06-2.0%)/0.5% (range, 0.11-2.2%) without/with risk factors.

CONCLUSIONS

Inverse RT plan optimization using bioeffect probability models allows individualization according to patient-specific risk factors. The modelled benefit when compared to clinical plans is, however, modest in most patients, demonstrating that current clinical plans are close to optimal. Larger gains may be achievable with morbidity endpoints rather than mortality.

Multi-GPU configuration of 4D intensity modulated radiation therapy inverse planning using global optimization

We report on the design, implementation and characterization of a multi-graphic processing unit (GPU) computational platform for higher-order optimization in radiotherapy treatment planning. In collaboration with a commercial vendor (Varian Medical Systems, Palo Alto, CA), a research prototype GPU-enabled Eclipse (V13.6) workstation was configured. The hardware consisted of dual 8-core Xeon processors, 256 GB RAM and four NVIDIA Tesla K80 general purpose GPUs. We demonstrate the utility of this platform for large radiotherapy optimization problems through the development and characterization of a parallelized particle swarm optimization (PSO) four dimensional (4D) intensity modulated radiation therapy (IMRT) technique. The PSO engine was coupled to the Eclipse treatment planning system via a vendor-provided scripting interface. Specific challenges addressed in this implementation were (i) data management and (ii) non-uniform memory access (NUMA). For the former, we alternated between parameters over which the computation process was parallelized. For the latter, we reduced the amount of data required to be transferred over the NUMA bridge. The datasets examined in this study were approximately 300 GB in size, including 4D computed tomography images, anatomical structure contours and dose deposition matrices. For evaluation, we created a 4D-IMRT treatment plan for one lung cancer patient and analyzed computation speed while varying several parameters (number of respiratory phases, GPUs, PSO particles, and data matrix sizes). The optimized 4D-IMRT plan enhanced sparing of organs at risk by an average reduction of [Formula: see text] in maximum dose, compared to the clinical optimized IMRT plan, where the internal target volume was used. We validated our computation time analyses in two additional cases. The computation speed in our implementation did not monotonically increase with the number of GPUs. The optimal number of GPUs (five, in our study) is directly related to the hardware specifications. The optimization process took 35 min using 50 PSO particles, 25 iterations and 5 GPUs.

Local Best Particle Swarm Optimization Using Crown Jewel Defense Strategy

Particle swarm optimization (PSO) is a swarm intelligence algorithm well known for its simplicity and high efficiency on various problems. Conventional PSO suffers from premature convergence due to the rapid convergence speed and lack of population diversity. It is easy to get trapped in local optima. For this reason, improvements are made to detect stagnation during the optimization and reactivate the swarm to search towards the global optimum. This chapter imposes the reflecting bound-handling scheme and von Neumann topology on PSO to increase the population diversity. A novel crown jewel defense (CJD) strategy is introduced to restart the swarm when it is trapped in a local optimum region. The resultant algorithm named LCJDPSO-rfl is tested on a group of unimodal and multimodal benchmark functions with rotation and shifting. Experimental results suggest that the LCJDPSO-rfl outperforms state-of-the-art PSO variants on most of the functions.

Inversed-Planned Respiratory Phase Gating in Lung Conformal Radiation Therapy

PURPOSE

To assess whether the optimal gating window for each beam during lung radiation therapy with respiratory gating will be dependent on a variety of patient-specific factors, such as tumor size and location and the extent of relative tumor and organ motion.

METHODS AND MATERIALS

To create optimal gating treatment plans, we started from an optimized clinical plan, created a plan per respiratory phase using the same beam arrangements, and used an inverse planning optimization approach to determine the optimal gating window for each beam and optimal beam weights (ie, monitor units). Two pieces of information were used for optimization: (1) the state of the anatomy at each phase, extracted from 4-dimensional computed tomography scans; and (2) the time spent in each state, estimated from a 2-minute monitoring of the patient's breathing motion. We retrospectively studied 15 lung cancer patients clinically treated by hypofractionated conformal radiation therapy, for whom 45 to 60 Gy was administered over 3 to 15 fractions using 7 to 13 beams. Mean gross tumor volume and respiratory-induced tumor motion were 82.5 cm³ and 1.0 cm, respectively.

RESULTS

Although patients spent most of their respiratory cycle in end-exhalation (EE), our optimal gating plans used EE for only 34% of the beams. Using optimal gating, maximum and mean doses to the esophagus, heart, and spinal cord were reduced by an average of 15% to 26%, and the beam-on times were reduced by an average of 23% compared with equivalent single-phase EE gated plans (P<.034, paired 2-tailed t test).

CONCLUSIONS

We introduce a personalized respiratory-gating technique in which inverse planning optimization is used to determine patient- and beam-specific gating phases toward enhancing dosimetric quality of radiation therapy treatment plans.