Machine Learning Dosimetry for Dynamic Radiopharmaceutical Therapy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radiopharmaceutical therapy dosimetry is a time-consuming manual process, and conventional imaging methods struggle to accurately localize small tumors, leading to inefficiencies in treatment planning and potential toxicity risks.
Innovation Solution
A machine learning system that generates predicted SPECT scans from PET and CT scans, automating dosimetry calculations and tumor tracking to optimize radiopharmaceutical therapy by predicting pharmacokinetics and adjusting therapy based on actual scans during the therapeutic phase.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual dosimetry calculation is used, then treatment planning can be performed, but the process is time-consuming and inefficient
Solution Approach 1:
The patent replaces the manual mechanical process of dosimetry calculation with an automated machine learning system. The ML model automatically segments tumors and organs from imaging data, predicts radiopharmaceutical uptake, and calculates dosimetry parameters, eliminating the need for manual intervention and significantly improving calculation efficiency while reducing time loss.
Solution Approach 2:
The system enables self-service dosimetry calculation by automatically processing imaging data and generating dosimetry reports without requiring manual radiologist intervention. The ML model performs tumor segmentation, organ segmentation, uptake prediction, and dose calculation in an autonomous manner, allowing the system to serve itself in completing the entire dosimetry workflow.
2Measurement precision
If conventional imaging methods are used, then basic imaging can be obtained, but small tumors cannot be accurately localized
Solution Approach 1:
The patent merges multiple imaging modalities (PET and CT scans) to overcome the limitations of conventional single-modality imaging. By combining the functional information from PET with the anatomical detail from CT, the system achieves superior tumor localization accuracy, particularly for small tumors that are difficult to detect with either modality alone.
Solution Approach 2:
The system performs preliminary tumor segmentation and identification from diagnostic imaging data before therapy administration. This preliminary action allows the ML model to establish baseline tumor characteristics and locations, which are then used to guide therapy planning and improve the detection and localization of small tumors during treatment.
3Adaptability or versatility
If fixed therapy dosage is administered, then treatment can be delivered, but individual patient response variations cannot be addressed
Solution Approach 1:
The patent implements dynamic therapy dosing by using the ML model to predict individual patient responses and adjust dosages accordingly. The system calculates personalized dosimetry values based on patient-specific factors including tumor characteristics, organ function, and predicted radiopharmaceutical uptake, enabling adaptive therapy that responds to individual patient variations while maintaining treatment reliability.
Solution Approach 2:
The system incorporates feedback mechanisms where dosimetry calculations from actual patient responses are used to refine and adjust subsequent therapy dosages. The ML model learns from patient-specific uptake patterns and treatment outcomes, continuously improving its predictions and enabling personalized dose optimization that addresses individual response variations while maintaining overall treatment consistency.
Data Source
AI summary
A system and method of using machine learning to predict the pharmacokinetics of a therapeutic radiopharmaceutical on a subject patient using the biodistribution data of the patient in order to dynamically treat the patient using the radiopharmaceutical.