Real-time tumor tracking with an artificial neural networks-based method: a feasibility study.

Area of Science:

• Medical Physics
• Radiation Oncology
• Artificial Intelligence

Background:

• Motion compensation strategies are crucial for accurate radiation therapy dose delivery.
• Current tumor tracking methods are limited in particle therapy due to stringent accuracy requirements.
• Artificial neural networks (ANNs) offer potential for advanced motion estimation.

Purpose of the Study:

• To develop and evaluate an ANN-based tumor tracking method for radiation therapy.
• To assess its performance in both photon and particle therapy applications.
• To improve the accuracy and feasibility of real-time tumor motion compensation.

Main Methods:

• Developed an ANN algorithm to estimate internal tumor trajectory from external surrogate signals.
• Retrospectively analyzed clinical data from 20 photon radiotherapy patients using infra-red motion tracking as a benchmark.
• Integrated the ANN into a hardware platform for particle therapy and tested on a moving phantom.

Main Results:

• Achieved a median tracking error reduction of up to 0.7 mm compared to state-of-the-art methods in clinical data.
• Demonstrated real-time tumor position estimation feasibility at 60 Hz acquisition rate in phantom studies.
• Validated the ANN's capability for high-accuracy tumor tracking.

Conclusions:

• Artificial neural networks are effective tools for developing high-accuracy, real-time tumor tracking systems.
• The proposed method enhances precision in radiation therapy, particularly for particle therapy.
• This technology has the potential to significantly improve treatment efficacy and patient outcomes.