Comparative effectiveness of standard vs. AI-assisted PET/CT reading workflow for pre-treatment lymphoma staging: a
Russell Frood1,2, Julien M Y Willaime3, Brad Miles4
1Department of Radiology, Leeds Teaching Hospitals NHS Trust, Leeds, United Kingdom.
Artificial intelligence (AI) in Fluorine-18 fluorodeoxyglucose (FDG)-positron emission tomography/computed tomography (PET/CT) reporting speeds up reads for experienced radiologists, increasing confidence without sacrificing quality. Less experienced readers were more affected by AI segmentation errors.
Area of Science:
- Radiology and Nuclear Medicine
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
Background:
- Fluorine-18 fluorodeoxyglucose (FDG)-positron emission tomography/computed tomography (PET/CT) is crucial for staging high-grade lymphoma.
- Reporting time for PET/CT scans varies based on case complexity.
- Integrating artificial intelligence (AI) into reporting workflows can enhance efficiency and quality.
Purpose of the Study:
- To assess the impact of an AI segmentation tool on the speed and quality of PET/CT reporting for lymphoma staging.
- To evaluate how AI assistance affects reader confidence and reporting behavior across different experience levels.
- To determine if AI integration influences the identification of disease sites in PET/CT scans.
Main Methods:
- A two-part reader study involving nine blinded reporters (trainees, junior, and senior consultants) evaluated 15 lymphoma staging PET/CT scans.
- Scans were reviewed first with a standard workflow and then with AI assistance incorporating pre-segmented disease sites.
- Report quality was independently assessed, and reader confidence was measured via questionnaires.
Main Results:
- AI-assisted reads showed a significant decrease in reporting time (median 13.3 min vs. 15.0 min, p < 0.001), particularly for junior and senior consultants.
- Report quality remained comparable between AI-assisted and standard workflows.
- AI assistance significantly increased reader confidence in disease identification (p < 0.001), though less experienced readers were more susceptible to segmentation errors.
Conclusions:
- AI-assisted PET/CT reading workflows can improve reporting efficiency without compromising quality.
- The AI tool demonstrated potential for cost reduction and faster report turnaround times.
- Further validation in larger studies is recommended, especially concerning the impact of segmentation errors on less experienced readers.
More Related Videos
07:15Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
07:57Positron Emission Tomography-based Dose Painting Radiation Therapy in a Glioblastoma Rat Model using the Small Animal Radiation Research Platform
Published on: March 24, 2022
