A tomographic reconstruction algorithm for cross-sectional imaging of IMRT beams from six projections.
Odran PivotPatrick PittetRolf ClackdoyleLaurent DesbatSimon RitPublished in: Physics in medicine and biology (2023)
Patient-specific Quality Assurance (QA) measurements are of key importance in radiotherapy for safe and efficient treatment delivery and allow early detection of clinically relevant errors. Such QA processes remain challenging to implement for complex IMRT radiotherapy fields delivered using a multileaf collimator (MLC) which often feature small open segments and raise QA issues similar to those encountered in small field dosimetry. Recently, detectors based on long scintillating fibers have been proposed to measure a few parallel projections of the irradiation field with good performance for small field dosimetry. The purpose of this work is to develop and validate a novel approach to reconstruct MLC-shaped small irradiation fields from six projections.

Approach: The proposed field reconstruction method uses a limited number of geometric parameters to model the irradiation field. These parameters are iteratively estimated with a steepest descent algorithm. The reconstruction method was first validated on simulated data. Real data were measured with a water-equivalent slab phantom equipped with a detector made of 6 scintillating-fiber ribbons placed at 1m from the source. A radiochromic film was used to acquire a reference measurement of a first dose distribution in the slab phantom at the same source-to-detector distance and the treatment planning system (TPS) provided the reference for another dose distribution. In addition, simulated errors introduced on the delivered dose, field location and field shape were used to evaluate the ability of the proposed method to efficiently identify a deviation between the planned and delivered treatments.

Main results: For a first small IMRT segment, 3%/3mm, 2%/2mm and 2%/1mm gamma analysis conducted between the reconstructed dose distribution and the dose measured with radiochromic film exhibited pass rates of 100%, 99.9% and 95.7%, respectively. For a second and smaller IMRT segment, the same gamma analysis performed between the reconstructed dose distribution and the reference provided by the TPS showed pass rates of 100%, 99.4% and 92.6% for the 3%/3mm, 2%/2mm and 2%/1mm gamma criteria, respectively. Gamma analysis of the simulated treatment delivery errors showed the ability of the reconstruction algorithm to detect a 3% deviation between the planned and delivered doses, as well as shifts lower than 7mm and 3mm when considering an individual leaf and a whole field shift, respectively.

Significance: The proposed method allows accurate tomographic reconstruction of IMRT segments by processing projections measured with six scintillating-fiber ribbons and is suitable for water-equivalent real-time small IMRT segments QA.
Keyphrases
- machine learning
- cross sectional
- deep learning
- early stage
- radiation induced
- radiation therapy
- computed tomography
- emergency department
- high resolution
- image quality
- patient safety
- artificial intelligence
- big data
- gold nanoparticles
- squamous cell carcinoma
- ionic liquid
- photodynamic therapy
- combination therapy
- replacement therapy
- data analysis