Trainable Weka segmentation tool machine learning-enabled segmentation on features of orthopantomograms

Abstract

<h3>Objective</h3> The aims of this research were to provide an in-depth analysis of the latest benchmarks in oral imaging by studying the segmentation of radiographs using Trainable Weka Segmentation and to accurately automate segmentation where it can be implemented on a large scale of clients in order to simplify radiological diagnosis. <h3>Study Design</h3> The experimentation was conducted by modifying open-source radiographs from the UFBA UESC DENTAL IMAGES dataset. In order to simulate realistic conditions such as noise affecting the regions of interest, panoramic radiographs were degraded and blurred with Gaussian noise. Accuracy was quantified by measuring the difference between the automated radiograph and the dentist-annotated image using MorphoLibJ. To ensure precision in results, automated predicted segmentations were observed by an oral maxillofacial radiologist and compared with the dentist-renditioning annotations of the panoramic images. <h3>Results</h3> The TWS classifier on radiographs with an average of 32 teeth and greater (Dice value of 0.66) and an average of less than 32 teeth (F1 score of 0.59) was significant. The calculated t-value for the Jaccard index was 2.78 and the t-value for the Dice score was 2.81. The results, considering the statistical scores, were due to the IV. The radiographs with 32 teeth and greater had higher IoU and F1 scores because there was less of a discrepancy in tooth alignment. <h3>Conclusions</h3> Segmentation of dental radiographs can be conducted by machine learning instead of manually. <b>Statement of Ethical Review</b> Ethical Review or exemption was not warranted for this study