Related Experiment Video
Updated: Jan 15, 2026
High-Resolution Cardiac Positron Emission Tomography/Computed Tomography for Small Animals
Published on: December 16, 2022
User-centric eXplainable AI criteria for implementing AI-based denoising in PET/CT.
M Champendal1, R T Ribeiro2, H Müller3
1HESAV School of Health Sciences - Vaud, HES-SO University of Applied Sciences and Arts Western Switzerland, Avenue de Beaumont 21, 1011 Lausanne, Switzerland; Faculty of Biology and Medicine, University of Lausanne, Unicentre, CH, 1015 Lausanne, Switzerland.
Radiographers require both simple, global explanations and detailed, case-specific insights from AI denoising tools in PET/CT. Tailored eXplainable Artificial Intelligence (XAI) features are crucial for trust and adoption in clinical practice.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Nuclear Medicine
Background:
- Clinical adoption of AI-based denoising in PET/CT requires transparent and trustworthy tools.
- Radiographer needs and workflow integration are key for successful implementation.
Purpose of the Study:
- To determine essential characteristics of eXplainable Artificial Intelligence (XAI) tools for AI-based denoising in PET/CT.
- To align XAI tool features with radiographer needs for clinical adoption.
Main Methods:
- Two focus groups with ten radiographers from nuclear medicine departments.
- Analysis of radiographer needs using matching/mismatching ground truth scenarios.
- Content analysis to identify desired XAI tool characteristics.
Main Results:
- Radiographers need two explanation levels: global summaries with confidence levels and detailed, case-specific insights.
- Key questions for AI understanding include 'what,' 'when,' and 'how.'
- Effective XAI tools should be easy, adaptable, user-friendly, and non-disruptive.
Conclusions:
- XAI tools must offer both workflow-efficient global summaries and in-depth case-specific details.
- Meeting these dual explanation needs fosters trust and integration of AI denoising in PET/CT.
- Adaptive XAI tools tailored to radiographers accelerate AI adoption in PET/CT imaging.
