-
Study Design and Data Collection
- The study protocol was approved by the Marmara University School of Medicine Non-Interventional Clinical Research Ethics Committee.
- Photographic images of buccal mucosa with healthy and oral lichen planus lesions were collected using the CranioCatch program.
- The dataset was divided into training, verification, and test sets for both healthy mucosa and mucosa with oral lichen planus lesions.
-
Identification of Disease Features
- Oral lichen planus exhibits clinical findings ranging from reticular white plaques to mucosal erythema, erosions, ulceration, and hyperkeratotic plaques.
-
Deep Convolutional Neural Network Architecture
- The deep learning process utilized the GoogleNet Inception V3 architecture implemented with the Tensorflow library.
- The architecture involved changing the filter size and grid size to improve efficiency in processing the images.
-
Results of the Deep Learning Algorithm
- The AI deep learning model correctly classified all test photos for both healthy and diseased mucosa.
- The accuracy of the model in distinguishing between normal buccal mucosa and oral lichen planus lesions was 100%.
-
Significance and Future Implications
- The study presented a novel deep learning algorithm for the classification of oral lichen planus lesions from photographic images in Turkey.
- The findings suggest that deep learning has the potential to address the challenge of diagnosing oral lichen planus accurately.
- Future AI developments in healthcare should prioritize human interests as a core goal, aiming to improve diagnostic abilities and patient outcomes.
Table of Contents
