Automated data extraction and report analysis in computer-aided radiology audit: practice implications from...

Area of Science:

• Medical Imaging
• Computational Pathology
• Paediatric Radiology

Background:

• Adherence to paediatric post-mortem magnetic resonance imaging (MRI) protocols is crucial for data quality.
• Assessing protocol adherence and image quality traditionally requires significant manual effort.

Purpose of the Study:

• To evaluate local departmental adherence to paediatric post-mortem MRI protocols.
• To implement and assess a customised automated computational approach for data collection and quality assessment.

Main Methods:

• A retrospective review of 460 paediatric post-mortem MRI examinations was conducted.
• A computer program was developed to process DICOM files and extract imaging sequence details.
• Natural language processing (NLP) was used to analyze reports and determine diagnostic image quality.

Main Results:

• The automated system extracted data from 80 GB of data within hours with 96.7% accuracy.
• Full protocol adherence was observed in 66.3% of cases, and abbreviated protocol in 30.4%.
• Overall diagnostic image quality was achieved in 91.9% of studies, with higher rates for the full protocol (97.7%) compared to the abbreviated protocol (79.3%).

Conclusions:

• An automated computational approach is effective for auditing paediatric post-mortem MRI protocols.
• This method streamlines data collection and quality assessment, facilitating future audits and quality improvements.
• The established automated workflow simplifies re-auditing after implementing changes.