Author: Lawrence T. Dauer, Yusuf Emre Erdi, Yiming Gao, Dustin W. Lynch, Usman Mahmood 👨🔬
Affiliation: Department of Medical Physics, Memorial Sloan Kettering Cancer Center 🌍
Purpose: Radiation dose to patients in CT examinations (CTDIvol, DLP) is tracked in dose monitoring solutions. Some solutions push for automatically estimating patient organ doses based on exam images. Such automation promises convenience, but errors can occur in the process. This study aims to identify errors by testing automated organ dose estimation functions under development in dose monitoring solutions and raise awareness regarding these errors.
Methods: Automatic organ dose estimation function of the dose monitoring solution integrated into hospital information system was under development and tested. Nine commonly used protocols in four body regions (brain, head-neck, chest, and abdomen-pelvis) were checked by randomly sampling eighty patient CT exams. Factors including patient centering, bowtie filter type, CTDIvol and DLP were recorded. If errors in automatic organ dose estimates were observed, they were recorded, categorized and analyzed for potential causes.
Results: Most investigated abdomen-pelvis exams (18/20) were misclassified as chest exams and only doses to organs in or near chest region were estimated. Astronomically high dose estimates (1E+28 mGy) were found in head-neck exams (6/20) due to a software bug in automatic computation process. Four of six investigated scanners did not properly set up automatic organ dose computation in brain exams resulting in no data in half of exams (10/20). Dose monitoring solution reported missing key parameters and could not provide estimates in two abdomen-pelvis exams and one brain exam. However, dose estimates comparable to literature were provided for all (20/20) chest exams.
Conclusion: Errors, software bugs and limitations were observed in the testing of automated organ dose estimation across protocols and scanners while literature-comparable estimates were observed in chest exams. Erroneous/Unrealistic dose estimates could skew dose statistics and harm protocol optimization if not corrected. Users should be aware of the errors and cautiously use dose monitoring solutions with automation capabilities.