[Paper Review] On the tradeoff between treatment time and plan quality in rotational arc radiation delivery

David Craft, Thomas Bortfeld|ArXiv.org|Oct 26, 2009

Advanced Radiotherapy Techniques5 references3 citations

TL;DR

This paper proposes a mixed-integer linear programming model that jointly optimizes treatment delivery time and plan quality in rotational arc therapy (VMAT), treating delivery modality as an optimization output rather than an input. It shows that when treatment time is constrained, VMAT outperforms IMRT in plan quality, while IMRT remains optimal when time is not limited, with the model providing tight bounds to estimate near-optimal solutions for clinical planning.

ABSTRACT

A new delivery option for cancer centers equipped with linear accelerators fitted with multi-leaf collimators (MLC) -- i.e. centers which can perform intensity modulated radiation therapy (IMRT) -- is rotational delivery. In rotational delivery, the beam is on while the gantry is rotating and the MLC leaves are moving, thus treating the patient more efficiently (regarding time) than in IMRT. A consideration that should be examined when evaluating this new type of delivery method is the tradeoff between treatment plan quality and delivery time: what do we sacrifice in terms of plan quality by guaranteeing a 2 minute delivery, for example? In this paper we examine this question by studying a simplified 2D phantom where leaf and gantry motion are directly included in the optimization model. We formulate the model as a linear mixed integer program. Because of the difficulty in solving to optimality, we employ additional models which allow us to trap the true optimal solution between upper and lower bounds. The lower bound solutions reveal which beam directions are most useful for treatment (i.e. where the gantry should slow down), and this information is used to produce deliverable solutions close to the lower bound solutions. For the phantom cases studied, we show that when time is not an issue, IMRT solutions are optimal, but when allowable treatment time is constrained sufficiently, rotational delivery is the preferred choice.

Motivation & Objective

• To investigate the tradeoff between treatment delivery time and plan quality in rotational arc therapy (VMAT) versus intensity-modulated radiation therapy (IMRT).
• To develop an optimization framework that does not predefine the delivery technique (IMRT or VMAT), but instead lets the optimizer choose based on time and dosimetric constraints.
• To provide tight upper and lower bounds on optimal plan quality under fixed treatment time, enabling assessment of solution quality for clinical planning.
• To identify beam angles and gantry motion patterns that maximize plan quality under time constraints, informing efficient VMAT delivery strategies.

Proposed method

• Formulates a linear mixed-integer program that models gantry rotation, leaf motion, and beam intensity modulation simultaneously, with time discretized into angular steps.
• Uses a lower-bound model (DAGU) that relaxes integrality constraints on leaf motion and allows arbitrary gantry speed to estimate the theoretical best plan quality for a given time.
• Employs a proximal point algorithm (PPA) model to generate deliverable VMAT plans close to the lower bound, ensuring feasibility under mechanical constraints.
• Applies a spatial and temporal discretization of gantry angles and leaf positions, with maximum leaf speed constrained to 0.5 cm/degree to reflect real linear accelerator dynamics.
• Uses a convex dose objective function and dose constraints to ensure clinical relevance, with the optimization minimizing dose deviation from prescribed targets.
• Compares solutions across time constraints (e.g., T=180, T=90) to assess the impact of time on plan quality and delivery technique selection.

Experimental results

Research questions

• RQ1 How does treatment time constraint affect the achievable plan quality in VMAT compared to IMRT?
• RQ2 Can an optimization model that does not predefine the delivery technique (IMRT or VMAT) identify the optimal balance between delivery time and plan quality?
• RQ3 What beam angles and gantry motion patterns are most beneficial for achieving high plan quality under time constraints?
• RQ4 How close can deliverable VMAT plans come to the theoretical optimal plan quality under fixed time limits?
• RQ5 Under what anatomical or dosimetric conditions does VMAT become preferable to IMRT due to time constraints?

Key findings

• When treatment time is not constrained, IMRT provides the optimal plan quality, as the optimizer selects the best beam angles and intensities.
• For a treatment time of T=180 time units, the VMAT plan achieved a dosimetric objective function value within 0.3% of the optimal IMRT solution, indicating near-optimality.
• At T=90 time units, the VMAT plan was 5% worse than the optimal IMRT solution, demonstrating a measurable tradeoff between time and quality.
• In non-convex dose distribution scenarios, the gantry slows down at specific angles to deliver higher modulation, indicating where beam delivery should be paused.
• In convex dose scenarios (e.g., pancreas phantom), the MLC remains closed for much of the arc, suggesting that VMAT can deliver high-quality plans with minimal leaf motion.
• The lower-bound model (DAGU) effectively identifies the most useful beam directions for modulation, which can guide VMAT optimization and beam angle selection.