A novel radiological software prototype for automatically detecting the inner ear and classifying normal from
Abdulrahman Alkojak Almansi1, Sima Sugarova2, Abdulrahman Alsanosi3
1University of Pecs, Faculty of Engineering and Information Technology, Institute of Information and Electrical Technology, Pecs, Hungary.
Computers in Biology and Medicine
|March 3, 2024
Summary
This study developed an AI tool to automatically detect inner ear malformations from CT scans. The software accurately classifies normal and abnormal anatomy, aiding in radiological diagnosis.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Inner ear malformations (IEM) require accurate diagnosis.
- Current diagnostic methods can be time-consuming.
Purpose of the Study:
- Develop and validate an automated software prototype.
- Read Digital Imaging and Communications in Medicine (DICOM) files.
- Classify normal and IEM from head computed tomography (CT) scans.
Main Methods:
- Retrospective analysis of 1200 inner ear CTs.
- Developed automated cropping algorithms for precise inner ear isolation.
- Trained and validated a deep learning convolutional neural network (DL CNN) model.
- Implemented a graphical user interface (GUI).
Main Results:
- Automated cropping achieved 92.25% accuracy.
- DL CNN model achieved an AUC of 0.86.
- AI model demonstrated accuracy (0.812), precision (0.791), recall (0.8), and F1-score (0.766).
Conclusions:
- Successfully developed an automated workflow for classifying inner ear anatomy.
- The tool shows potential for risk stratification in radiological diagnosis.
- Clinical decision-making requires supervision by qualified medical professionals.
