A conjugate gradient-assisted multi-objective evolutionary algorithm for fluence map optimization in radiotherapy treatment

Ruifen Cao, Langchun Si, Xuesong Li, Yaopei Guang, Chao Wang, Ye Tian, Xi Pei, Xingyi ZhangComplex 2022 · DOI

Abstract

Intensity-modulated radiotherapy (IMRT) is a cornerstone of modern cancer treatment. A critical component of IMRT planning is fluence map optimization (FMO), which determines the intensity of radiation beams to maximize tumor destruction while minimizing damage to healthy tissue. Traditionally, this inverse problem is solved by manually adjusting weights in a single-objective function—a process that is both time-consuming and dependent on expert experience. This paper introduces a conjugate gradient-assisted multi-objective evolutionary algorithm (CG-MOEA) that reformulates FMO into a three-objective problem. By combining the global search of evolutionary algorithms with the local efficiency of conjugate gradient methods, CG-MOEA generates a diverse set of high-quality treatment plans in a single run, outperforming state-of-the-art methods across multiple clinical cases.

Introduction

In Intensity-Modulated Radiotherapy (IMRT), the primary goal is to deliver a lethal dose of radiation to the target tumor while sparing surrounding organs at risk (OARs). This is achieved through the Fluence Map Optimization Problem (FMOP), where the radiation beam is divided into a grid of "beamlets," each with a specific intensity or "fluence."

A grid-based fluence map representing beamlet intensities.

Optimizing these maps is inherently a multi-objective challenge. Clinicians must balance sufficient tumor coverage against the safety of adjacent healthy organs.

The Limitations of Traditional Planning

Current clinical practice often treats FMOP as a single-objective optimization problem by combining various constraints (e.g., maximum dose to the spinal cord, minimum dose to the tumor) into a single weighted sum. This approach has two major drawbacks:

  1. Manual Weight Tuning: Radiation physicists must iteratively adjust coupling weights based on trial and error, which is inefficient and subjective.
  2. Single Solution: The process typically yields only one plan. If that plan is clinically unsatisfactory, the entire optimization must be restarted with new weights.

The CG-MOEA Approach

To overcome these hurdles, the researchers developed the Conjugate Gradient-assisted Multi-Objective Evolutionary Algorithm (CG-MOEA). This framework introduces three key innovations:

1. Three-Objective Reformulation

Instead of managing dozens of individual constraints, CG-MOEA aggregates clinical requirements into three primary objectives: minimizing under-dosage, minimizing over-dosage, and maximizing dose uniformity. This reduction in dimensionality allows multi-objective evolutionary algorithms (MOEAs) to operate effectively without the "curse of dimensionality."

2. Hybrid Search Strategy

The algorithm utilizes a population-based search to explore a wide variety of potential plans. To accelerate convergence, it incorporates a Conjugate Gradient (CG) local search. While evolutionary algorithms are excellent at global exploration, they can be slow to fine-tune solutions. The CG method uses the mathematical gradient of the dose functions to rapidly push promising solutions toward the optimal "Pareto front."

A typical radiotherapy treatment setup using a linear accelerator (LINAC).

A common risk in hybrid algorithms is "premature convergence," where the local search forces the entire population to become too similar. CG-MOEA employs an adaptive strategy that monitors population diversity and adjusts the frequency of the CG local search accordingly, ensuring a diverse range of trade-offs for the clinician to consider.

Clinical Performance and Results

The CG-MOEA was tested on five real-world clinical cases, including nasopharyngeal, lung, cervical, esophageal, and kidney cancers. The algorithm was compared against five state-of-the-art MOEAs (such as NSGA-II and MOEA/D).

Key findings include:

  • Superior Convergence: CG-MOEA consistently achieved the highest Hypervolume (HV) values, a metric indicating both the quality and diversity of the solution set.
  • Clinical Excellence: Evaluation via Dose-Volume Histograms (DVH)—the clinical standard—showed that CG-MOEA produced plans that met or exceeded all medical requirements more consistently than traditional methods.
  • Efficiency: By providing a full Pareto set of optimal plans in a single run, the algorithm eliminates the need for manual weight-tweaking, potentially saving hours of planning time per patient.

Conclusion

CG-MOEA represents a significant step toward the automation of radiotherapy planning. By treating fluence map optimization as a true multi-objective problem and leveraging the synergy between evolutionary and gradient-based search, the algorithm provides clinicians with a powerful tool to find the best possible balance between cancer treatment and patient safety.

Summary updated

Abstract

Abstract Intensity-modulated radiotherapy (IMRT) is one of the most applied techniques for cancer radiotherapy treatment. The fluence map optimization is an essential part of IMRT plan designing, which has a significant impact on the radiotherapy treatment effect. In fact, the treatment planing of IMRT is an inverse multi-objective optimization problem. Existing approaches of solving the fluence map optimization problem (FMOP) obtain a satisfied treatment plan via trying different coupling weights, the optimization process needs to be conducted many times and the coupling weight setting is completely based on the experience of a radiation physicist. For fast obtaining diverse high-quality radiotherapy plans, this paper formulates the FMOP into a three-objective optimization problem, and proposes a conjugate gradient-assisted multi-objective evolutionary algorithm (CG-MOEA) to solve it. The proposed algorithm does not need to set the coupling weights and can produce the diverse radiotherapy plans within a single run. Moreover, the convergence speed is further accelerated by an adaptive local search strategy based on the conjugate-gradient method. Compared with five state-of-the-art multi-objective evolutionary algorithms (MOEAs), the proposed CG-MOEA can obtain the best hypervolume (HV) values and dose–volume histogram (DVH) performance on five clinical cases in cancer radiotherapy. Moreover, the proposed algorithm not only obtains the more optimal solution than traditional method used to solve the FMOP, but also can find diverse Pareto solution set which can be provided to radiation physicist to select the best treatment plan. The proposed algorithm outperforms dose-volume histogram state-of-the-art multi-objective evolutionary algorithms and traditional method for FMOP on five clinical cases in cancer radiotherapy.
BibTeX
@inproceedings{cao2022conjugate,
  author = {Ruifen Cao and Langchun Si and Xuesong Li and Yaopei Guang and Chao Wang and Ye Tian and Xi Pei and Xingyi Zhang},
  title = {{A conjugate gradient-assisted multi-objective evolutionary algorithm for fluence map optimization in radiotherapy treatment}},
  booktitle = {Complex},
  year = {2022},
  doi = {10.1007/s40747-022-00697-7},
}
Figure