Organ Dose Estimation in CT through Voxelized Phantoms and Monte Carlo Methods: Impact of Wrong Patient-Phantom Matching
PURPOSE The aim of this study was to assess the impact on organ dose estimation for Body CT examinations when the patient is matched to the wrong virtual phantom. METHODS AND MATERIALS Patient-specific organ dosimetry is estimated using a Monte Carlo framework and matching patients to voxelized virtual human phantoms (hereafter 'phantoms'). Software for automatic patient scout segmentation was developed to identify anatomical landmarks, which are then used to match the patient with a phantom model. From the AP scout collected by a dose tracking system (DoseWatch, GE Healthcare), the patient contour, gray scale intensity profiles, and bone symmetry and edges are used to identify anatomical regions. The extracted patient-specific landmarks were then used to select the matching phantom from a database of 69 phantoms. Patient-specific organ doses were automatically estimated for a sample of 130 patients who underwent a chest/abdomen/pelvis exam. To verify the accuracy of the software, the AP scouts were also visually inspected to identify the correct landmarks manually. These correct landmarks led to a matched phantom, which could be the same or a different one from the one enabled by the automatic landmarks detection. Organ doses were estimated again and were compared to the one estimated using the automatic software for landmarks recognition. RESULTS Preliminary results on 40 scouts show that the anatomical landmarks tool failed in detecting the correct landmarks in 17.5% of the cases (2% in chest, CAP, and kidney to bladder; 5% in abdomen and 5% in abdomen-pelvis) with an error ranging from 21 to 34 mm for abdomen, from 35 mm to 67 mm for abdomen-pelvis and up to 87 mm for CAP. The differences resulted in a wrong phantom selection in 15% of the cases. The associated differences in organ dose estimation due were higher for the partially irradiated organs (up to 96%) for all anatomical regions; for totally irradiated organ like stomach and liver the differences were up to 30%. CONCLUSION The study showed that error in dose estimation as large as 30% for fully irradiated may occur if the patient is matched to the wrong phantom. CLINICAL RELEVANCE/APPLICATION Quantifying and monitoring computed tomography (CT) radiation dose for an individual patient has become necessary to practice medical imaging risk assessment and protocol optimization. For this, an accurate organ dose estimation is needed.
