BEST IN PHYSICS (IMAGING): Automated Measurement and Monitoring of Patient-Specific Model-Observer-Based Lesion Detectability In CT Exams
Abstract
Purpose
Measuring and monitoring dose and noise for each patient exam has been proposed as part of a quality program for routine CT practice. Mathematical model observers have been recognized as more meaningful methods than noise level alone for image quality assessment in CT. However, its implementation on each patient’s CT exam is still limited. In this work, we implemented a software platform that automates patient-specific channelized Hotelling observer (CHO)-based analysis with integrated reporting of image quality and dose metrics.
Methods
A patient-specific CHO-based method to estimate lesion detectability for each individual patient scan was developed previously in our group. To implement this method in routine CT workflow, we developed a software using an ORTHANC DICOM server for image management and analysis, PostgreSQL for data storage, and a React-based dashboard for visualization. The system allows bulk DICOM pulls during server downtime to reduce clinical bottlenecks, then automatically extracts DICOM metadata to calculate size-specific dose estimates (SSDE), water equivalent diameter (Dw), and CTDIvol along the scan axis. Noise levels are determined from standard deviation histograms and noise power spectrum (NPS) is computed from background ROIs. The CHO analysis is based on configurable lesion models with contrast-dependent MTF. Detectability index (d') is calculated at configurable z-axis intervals.
Results
The platform processes patient CT images from multiple scanner models. Reports on noise level, SSDE, and d’, both globally and as a function of z coordinate, for each image series were generated. Users can filter results by patient, institution, scanner model, protocol, and date range. Results are stored with export capabilities for external analysis.
Conclusion
An automated, patient-specific CHO software was implemented to enable routine task-based quantitative assessment of lesion detectability for CT exams. This approach provides a scalable mechanism for objective quality monitoring and facilitates routine reporting of clinically meaningful image quality metrics.