The Use of Laboratory Tests in Diagnosing Lesions of The Mouth and Jaws

Abstract

Background: Determining odontogenic cysts and tumors necessitates an early diagnosis to avert surgeries involving extensive elimination of infected tissues. This study assesses the accuracy of the YOLO v2 deep learning network in contrast to conventional methods to detect dental caries in panoramic radiographs.

Methods: Research was done using 1602 lesions in periapical radiographs taken at Yonsei University Dental Hospital between the years 2010 and 2019. The study divided the participants into those with dentigerous cysts, odontogenic keratocysts, ameloblastoma, and the control group without any lesion. This paper aims to assess the diagnostic performance of YOLO v against that of oral and maxillofacial surgeons and general practitioners by using measures like precision, recall, accuracy, and the F1 score that will be used for objective evaluation. 

 Results: YOLO showed the highest metric accuracy among the three teams, with particularity percentage and amount of recall being 0.707 and 0.680, respectively. Although the results of the YOLO models did not vary much from clinical performance, none of these differences were statistically observed.

Discussion: The results of this study may point towards the possibility of the YOLO version becoming successful in the detection of jaw cysts and tumors in panoramic radiography tests. It is as competent as a human clinician in this case, which makes a potential introduction either an aid to screen patients early and curb those unnecessary morbidities in oral and maxillofacial surgery.

Conclusions: The study stresses the benefits that auto-detecting machine learning algorithms like YOLO offer in medical process automation and AI in dentistry. The good performance of YOLO in hunting for lesions and creating pictures on panoramic radiographs shows its ability to facilitate diagnosis and thus influence positive patient outcomes.