Dynamic modulated arc therapy (DMAT): A time-aware, modulation-steered optimization framework for next-generation...

Conventional VMAT optimization largely treats delivery time and deliverability as emergent properties of simplified, control-point-centric models that often ignore finite acceleration and other dynamic limits. As modern linacs continue to increase maximum axis speeds and dose rates, there is a growing need for a planning paradigm that makes the plan quality-time trade-off explicit and steerable, especially when shorter treatments can provide clinical benefits beyond throughput.

This work introduces Dynamic Modulated Arc Therapy (DMAT), a time-aware, modulation-steered framework that jointly optimizes dosimetric quality, delivery time, and modulation complexity. Here, time-aware denotes that delivery time is computed within the optimizer from an explicit model of machine dynamics, including finite acceleration and axis synchronization, rather than inferred after optimization; modulation-steered denotes that leaf-travel allowance, control-point density, total monitor units (MUs), and aperture complexity are governed together by a single user-selected modulation level, making the amount and distribution of modulation an explicit input to the optimizer. Clinical goals enter the optimizer as metric-based cost functions defined by each criterion's metric, goal value, acceptable variation, and priority, with supporting objectives generated automatically.

DMAT couples (1) direct machine emulation that accounts for axis synchronization and dynamic limits, (2) dynamic modulation control, and (3) clinical metrics as optimization cost functions. A user-selected modulation level (-3 to +3) governs leaf-travel allowance, total MU behavior, aperture-complexity penalties, and control-point (CP) density. Plans are created by defining a treatment trajectory, initializing CP geometry, leaf positions, and MU, then iteratively improving the solution with a progressive-resolution loop alternating dosimetric updates with sequencing/deliverability updates, followed by post-processing to reduce complexity with minimal dose change. CP density is adapted nonuniformly: a preliminary uniform-CP optimization is followed by redistribution according to combined delivery-complexity and geometric information. DMAT was evaluated using a hypothetical accelerated C-arm delivery system (max gantry speed 2.5 RPM, max MLC speed 6.25 cm/s, max dose rate 3000 MU/min) on representative head-and-neck, lung SBRT, and prostate SBRT cases.

Across all cases, increasing modulation level produced plans with increased modulation surrogates (e.g., MU/Gy and aperture complexity) and progressively longer delivery time, while adaptive non-uniform CP allocation concentrated additional CPs in arc sectors where higher angular resolution was most beneficial. The resulting quality-time trade-off space was disease site dependent: the head-and-neck case showed substantial plan-quality gains with increased modulation and CP density, whereas prostate SBRT and lung SBRT exhibited smaller incremental quality improvements beyond baseline despite longer delivery times. When efficiency was prioritized (negative modulation levels), DMAT predictably reduced modulation and maintained a constant CP budget while shortening delivery time, producing corresponding and quantifiable reductions in plan quality.

DMAT establishes a time-aware, modulation-steered planning and delivery methodology in which plan quality and modulation complexity are co-optimized under explicit clinical and user control using machine-aware timing. Accurate delivery time is exposed to users during planning through advanced machine emulation. By doing so, DMAT makes quality-time trade-offs transparent, predictable, and navigable, providing a practical foundation for leveraging next-generation delivery systems' capabilities and supporting time-constrained workflows such as motion-sensitive treatments and adaptive radiotherapy.