Radiomics-Based Diagnosis in Dentomaxillofacial Radiology: A Systematic Review.

Area of Science:

• Dentomaxillofacial Radiology
• Quantitative Imaging Analysis
• Radiomics

Background:

• Radiomics offers advanced quantitative analysis of medical images.
• Its application in Dentomaxillofacial Radiology (DMFR) is an emerging field.
• Systematic reviews are crucial for evaluating emerging technologies in specialized radiology.

Purpose of the Study:

• To systematically review and evaluate the scientific literature on Radiomics applications in Dentomaxillofacial Radiology.
• To assess the potential implications and accuracy of Radiomics in diagnosing DMFR pathologies.
• To identify common Radiomics modeling types and feature statistics used in DMFR.

Main Methods:

• A systematic literature search was conducted, including studies on human samples and in vivo DMFR applications.
• Fifty-one articles published from 2015 onwards were selected from 3789 initial records.
• The QUADAS-2 tool assessed the risk of bias, with prediction accuracy and modeling type as primary and secondary outcomes.

Main Results:

• Selected studies reported prediction accuracies for DMFR pathologies ranging from 0.66 to 99.65%.
• Logistic regression was the most frequent Radiomics modeling technique, followed by Support Vector Machine and Decision Tree.
• Second-order statistics were the most commonly applied Radiomics features, followed by first-order, higher-order, and shape-based statistics.

Conclusions:

• Radiomics demonstrates significant potential as a quantitative tool in Dentomaxillofacial Radiology, offering an alternative to conventional visual assessment.
• Further standardization of the Radiomics workflow is essential for consistent and reliable results.
• While Radiomics alone doesn't drive digital transformation, its integration into digital workflows holds considerable promise for DMFR.