Deep Learning-Based Software Improves Clinicians' Detection Sensitivity of Aneurysms on Brain TOF-MRA.
B Sohn1, K-Y Park2, J Choi1,3,4
1From the Department of Radiology (B.S., J.C., J.H.K., K.H., B.J., S.Y.W., J.C., H.S.C., S.-K.L.), Research Institute of Radiological Science and Center for Clinical Imaging Data Science, Yonsei University College of Medicine, Seoul, South Korea.
AJNR. American Journal of Neuroradiology
|August 13, 2021
Summary
Deep learning software significantly improved the detection of cerebral aneurysms on MRA for neurosurgeons and neurologists. This AI tool enhanced sensitivity without substantially increasing false positives, aiding clinical diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Neurology
Background:
- Cerebral aneurysm detection on Magnetic Resonance Angiography (MRA) is challenging.
- Deep learning (DL) software shows promise for automated aneurysm detection on MRA.
- Evaluating DL software's impact on clinical interpretation is crucial.
Purpose of the Study:
- To assess the added value of DL software in detecting cerebral aneurysms on MRA.
- To compare diagnostic performance with and without DL software assistance.
- To evaluate the impact on sensitivity and false-positive rates among different medical specialists.
Main Methods:
- Retrospective analysis of 332 time-of-flight MRA (TOF-MRA) examinations for intracranial aneurysms.
- Four physicians (2 radiologists, 1 neurosurgeon, 1 neurologist) interpreted MRA blindly, then with DL software assistance.
- Key metrics included per-patient sensitivity, per-lesion sensitivity, specificity, and false-positive rates.
Main Results:
- Software assistance significantly improved patient-based sensitivity from 73.5% to 86.5% (P < .001).
- Neurosurgeons and neurologists showed significant sensitivity increases (e.g., neurologist: 56.3% to 84.4%, P < .001).
- False-positive rates did not significantly increase with software use.
Conclusions:
- DL software significantly enhances clinician sensitivity for MRA aneurysm detection, particularly for neurosurgeons and neurologists.
- The incremental value of DL software was equivocal for radiologists in this study.
- AI-assisted interpretation offers a valuable tool for improving aneurysm detection accuracy without compromising specificity.
