Congress:
EuroSafe Imaging 2020
Keywords:
Multicentre study, Observational, Retrospective, Quality assurance, Education and training, Safety, Health policy and practice, Dosimetry, CT, Radioprotection / Radiation dose, Radiographers, Action 5 - Performance indicators for radiation protection management
Authors:
T. Szczykutowicz, S. Rose
DOI:
10.26044/esi2020/ESI-02989
Description of activity and work performed
We built a pattern recognition algorithm to automatically detect repeats using an informatics solution.
- IT solution automatically calculates repeat rate
- IT solution requires no manual identification of repeats by technologists or radiologists
- IT solutions learns from a site's own practice patterns, making it capable of be deployed across a center with many different locations
- The method uses standard DICOM metadata, direct vendor enabled connection to modality units is therefore not required. Our method requires: scan mode (i.e., axial/sequential, helical/spiral, scout/topogram, etc.), scanner protocol name, accession number, starting slice location (e.g., "inferior 50 mm"), end slice location (e.g., "superior 23 mm"), series description, and study description/orderable.
- We provide an overview of the algorithm in Figure 3.
Inspired by our earlier work on identifying "rogue" technologists who decided at scan time to not adhere to approved protocols and independently made changes, we used a pattern recognition algorithm to perform repeat rate analysis.[4]
- We determine whether an exam contains an acquisition that overlaps with another scan in the same exam using the start and end slice locations from each scan event.
- Then we use the scan mode and scanner protocol name from every scan event within an exam to form a digital signature for individual exams. The signature includes the quantity of overlapped scan events that occurred.
- These signatures are fed into a pattern recognition algorithm, which calculates the frequency of every type of scan signature at an institution. We identify something we call a base signature, which is the "gold standard" way a given exam should be performed. Gold standard scan signatures are those that occur most commonly, with the assumption that the majority of a site’s scans hopefully do not contain repeats.
- After we associate each exam with a base signature, we compare them. When a signature differs from the gold standard, it is usually due to a repeat. We classify the repeats into the following types based on their scan mode: helical/spiral overlap, axial/sequential overlap, helical/spiral extension, axial/sequential extension, bolus-tracking, and localizer.
At the end of the day, our method classifies every exam as either containing a repeat or not, and indicates the type of repeat that was scanned to facilitate additional training or workflow modifications. Since we also know the site’s locations and performing technologist, we can perform targeted remediation.
We applied our repeat rate algorithm to 2 different sites. Site A: rural hospital and Site B: academic hospital. We included data from 3 years time, representing 61,102 patient exams.
We identified the exams with the highest repeat rates for the helical and axial scan modes at Site A and B. We then calculated the effective dose for these exams, separately for exams not containing and containing a repeat.