Automated batch dosimetry data processing software for improved clinical applications and workflow

Objectives: Personalized radionuclide dosimetry offers the potential of reducing risk of high dose to vulnerable normal organs while maximizing tumor dose. However, dosimetry entails significant time/efforts. Improving workflows to be more robust, dependable and faster may aid in bringing them into routine clinical operations. In this study, we created dose calculation workflow for Lu-177 DOTATATE dosimetry that integrates vendor image analysis software with in-house software for curve fitting, biodistribution modelling, and dosimetry calculations within an automated workflow and database architecture. Methods: We built in-house software (in IDL) entitled Dosimetry Data Processor (DDP), to support a nuclear medicine physics dose calculation workflow. The workflow has 2 main components: (1) organ uptake characterization via image analysis using vendor software and (2) automated dose processing including curve fitting, numerical integration, error checking, and absorbed dose calculation (using OLINDA/EXM v1.1 Dose Conversion Factors). In addition, all DDP input is automatically archived, enabling absorbed dose to be calculated (or recalculated) via batch data processing. To validate the system, dosimetry calculations were compared with standard OLINDA/EXM software (v1.1) in 5 test cases, and evaluated for speed and ease of use for sets of post-therapy images following 199 treatments. To demonstrate the ease/benefit and versatility of our batch reprocessing capacity, population dosimetry was calculated using both mono-exponential curve fitting and trapezoidal integration/physical half-life decay modelling assumptions. RESULTS: In 5 test cases, dose calculations using DDP yielded results identical to OLINDA/EXM. Median physicist time required to perform dosimetry using DDP in routine cases (i.e. after imaging/blood data available) was approximately 1 hr per patient/treatment cycle (~45 min image analysis, 10 min blood measurement verification and formatting, 5 min processing using DDP). Once a patient's biodistribution details were logged into the database, dosimetry calculation required approximately 0.6 seconds and could be run automatically for an entire population. To demonstrate the ability to test the effect of our assumptions on the population dosimetry, we compared curve fitting methodologies in our population. We found that the dose to the kidneys, often the dose limiting organ, varied significantly between exponential curve fitting models and trapezo