ML-EM algorithm for dose estimation using PET in proton therapy.

Area of Science:

• Medical Physics
• Nuclear Medicine
• Radiation Oncology

Background:

• Positron emission tomography (PET) is investigated for proton therapy dose verification.
• Current PET methods face limitations as positron emitter distributions do not directly correlate with dose distributions.
• Accurate in vivo dose verification is crucial for effective proton therapy.

Purpose of the Study:

• To develop and validate a novel method for estimating proton therapy dose distributions from PET data.
• To improve the accuracy and speed of dose verification in proton therapy using advanced algorithms.
• To enable real-time dose monitoring during proton therapy treatments.

Main Methods:

• Utilized the maximum likelihood (ML) expectation maximization (EM) algorithm combined with filtering for dose estimation.
• Irradiated a water target with mono-energetic and spread-out Bragg-peak proton beams.
• Performed planar PET measurements and Monte Carlo simulations for filter function derivation and analysis.

Main Results:

• Successfully estimated 2D dose distributions from PET images with approximately 10% relative error, comparable to PET activity measurement deviations.
• Achieved low error (<5%) for laterally integrated profiles along the beam direction.
• Estimated proton ranges within 1 mm and completed 2D dose estimation in 21 ms, demonstrating potential for real-time monitoring.

Conclusions:

• The proposed ML-EM algorithm with filtering provides accurate and fast estimation of proton dose distributions from PET data.
• This method overcomes limitations of conventional PET-based dose verification, enabling better in vivo accuracy.
• The algorithm represents a significant advancement towards exploiting PET for real-time, in vivo dose verification in proton therapy.