Differential Diagnosis of OKC and SBC on Panoramic Radiographs: Leveraging Deep Learning Algorithms

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Odontogenic keratocyst (OKC) and simple bone cyst (SBC) are common jaw cysts with distinct clinical and histological features.
• Accurate preoperative differentiation between OKC and SBC is crucial for appropriate surgical management and prognosis.
• Panoramic radiography is a widely used imaging modality in dentistry, but differentiating these cysts can be challenging based on radiographic features alone.

Purpose of the Study:

• To evaluate the efficacy of a deep learning algorithm in distinguishing OKC from SBC using only preoperative panoramic radiographs.
• To assess the diagnostic performance of the AI model in classifying these two cyst types.

Main Methods:

• A retrospective analysis of 63 histologically confirmed OKC cases and 125 SBC cases was performed.
• Panoramic radiographs were processed, cropped to 299x299 images, and divided into training (80%) and validation (20%) sets.
• The Inception-ResNet-V2 deep learning model was employed for OKC and SBC discrimination with 5-fold cross-validation.

Main Results:

• The deep learning model achieved an accuracy of 0.829, precision of 0.800, recall of 0.615, and an F1 score of 0.695.
• Class activation mapping (CAM) visualization facilitated the understanding of the algorithm's decision-making process.

Conclusions:

• The deep learning algorithm demonstrates significant potential for accurately differentiating OKC from SBC on panoramic radiographs.
• This AI-driven approach can serve as a valuable adjunct for clinicians, potentially enhancing diagnostic accuracy and guiding treatment strategies.
• Further validation and integration into clinical workflows could improve patient outcomes by enabling more precise diagnoses.