Related Experiment Video
Updated: Jun 14, 2025
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.3K
Digital Otoscopy With Computer-Aided Composite Image Generation: Impact on the Correct Diagnosis, Confidence, and
Seda Camalan1, Carl D Langefeld2, Amy Zinnia2
1Center for Artificial Intelligence Research, Wake Forest University School of Medicine, Winston-Salem, North Carolina, USA.
Summary
Ear, nose, and throat (ENT) physicians performed similarly using computer-assisted images (SelectStitch) and single frames (Still) for detecting ear abnormalities. Video clips improved accuracy but increased diagnosis time, highlighting AI potential in otoscopy.
Area of Science:
- Otolaryngology
- Medical Imaging
- Artificial Intelligence
Background:
- Digital otoscopy is crucial for diagnosing ear conditions.
- Various visualization methods exist for interpreting otoscopic images, each with potential benefits and drawbacks.
- Computer-assisted image generation aims to improve diagnostic efficiency and accuracy.
Purpose of the Study:
- To compare the diagnostic performance of ear, nose, and throat (ENT) physicians using three digital otoscopy visualization methods: SelectStitch (AI-composite images), Still (single video frame), and Video (full clip).
- To evaluate clinicians' diagnostic confidence levels and the time required for diagnosis across the different methods.
- To explore the utility of AI-driven techniques in otoscopic diagnosis.
Main Methods:
- A clinician diagnostic reader study involving nine ENT physicians.
- Physicians reviewed digital otoscopy examinations from 86 ears with diverse pathologies.
- Examinations were presented using SelectStitch, Still, or Video visualization methods.
Main Results:
- Physician ability to detect ear abnormalities ranged from 33.2% to 68.7%, varying by pathology.
- SelectStitch and Still images showed no statistically significant difference in detection rates.
- Video clips yielded significantly better detection rates (P < .01) but required substantially longer diagnosis times compared to SelectStitch and Still (P > .05).
- Diagnostic confidence correlated positively with correct diagnoses but varied by specific ear pathology.
Conclusions:
- Computer-assisted techniques like SelectStitch show promise for enhancing otoscopic diagnoses and saving time, particularly beneficial for telemedicine.
- The comparable performance of SelectStitch and manually selected Still images suggests the potential of AI algorithms in otoscopy.
- While video clips offer improved accuracy, the trade-off in time necessitates further investigation into optimizing AI-driven visualization for clinical efficiency.
