Article #2

Introduction

• Study Purpose: Investigate AI-powered clinical decision support (CDS) for radiology trainee education.

• Focus: Provide automated real-time feedback during the interpretation of clinical and simulation brain MRI examinations.

Methods

• Participants: Neuroradiology fellows and diagnostic radiology residents.

• Study Design: Each trainee interpreted 3 brain MRIs in reading sessions:

  • Two clinical cases (one with CDS, one without).

  • One teaching file-based simulation case with CDS.

• CDS System Details:

  • Bayesian inference-based system.

  • Trainees input imaging features and differential diagnoses.

  • Software provides inferred diagnoses based on input.

  • Supports educational resources linked to differential diagnosis.

Results

• Sessions Conducted:

  • 75 brain MRI examinations reviewed across 25 sessions with 10 trainees.

• Confidence Ratings:

  • Trainees had lower confidence with simulation cases that used CDS compared to clinical cases without CDS (p<0.05).

  • Higher educational value rated for CDS simulation cases over traditional clinical cases.

• Timing Analysis:

  • No significant differences in overall time taken for different case scenarios.

  • Simulation cases showed reduced time for clinical correlation in posture (p < 0.05).

Findings

• Educational Value:

  • Simulation cases provided an enhanced educational experience, particularly for less common lesions.

• Feedback Impact:

  • Trainees adjusted their differential diagnoses upon seeing CDS results in 44% of cases.

  • Utilization of educational links was minimal (4%).

Discussion

• Key Findings:

  • CDS enhances educational value for complex cases.

  • Limitations in current clinical case teaching due to variability in case exposure and teaching styles.

• Technology Implications:

  • CDs can supplement traditional learning and improve exposure to diverse pathologies in a controlled environment.

Limitations

• Sample Size: Small number of participants may limit generalizability.

• Potential Bias: Participants aware of simulation vs. clinical case distinction may affect results.

• Future Research: Explore integration of more inpatient/emergency cases.

Conclusion

• Recommendation: Incorporating simulation training and automated feedback in radiology education may improve trainee learning and adaptation to various clinical scenarios.

References

• Extensive references list providing background on simulation in radiology and AI in clinical education.