Rafał Obuchowicz1, Karolina Nurzynska2, Barbara Obuchowicz3, Andrzej Urbanik4, Adam Piórkowski5. 1. Department of Diagnostic Imaging, Jagiellonian University Medical College, 19 Kopernika Street, 31-501, Cracow, Poland. rafalobuchowicz@su.krakow.pl. 2. Institute of Informatics, Faculty of Automata Control, Electronics, and Computer Science, Silesian University of Technology, Akademicka 16, 44-100, Gliwice, Poland. 3. Department of Conservative Dentistry with Endodontics, Jagiellonian University Collegium Medicum, Montelupich 4, 31-155, Cracow, Poland. 4. Department of Diagnostic Imaging, Jagiellonian University Medical College, 19 Kopernika Street, 31-501, Cracow, Poland. 5. Department of Geoinformatics and Applied Computer Science, AGH University of Science and Technology, Mickiewicza 30, 30-059, Cracow, Poland.
Abstract
OBJECTIVES: Dental caries are caused by tooth demineralization due to bacterial plaque formation. However, the resulting lesions are often discrete and thus barely recognizable in intraoral radiography images. Therefore, more advanced detection techniques are in great demand among dentists and radiographers. This study was performed to evaluate the performance of texture feature maps in the recognition of discrete demineralization related to caries plaque formation. METHODS: Digital intraoral radiology image analysis protocols incorporating first-order features (FOF), co-occurrence matrices, gray tone difference matrices, run-length matrices (RLM), local binary patterns (LBP), and k-means clustering (CLU) were used to transform the digital intraoral radiology images of 10 patients with confirmed caries, which were retrospectively reviewed in a dental clinic. The performance of the resulting texture feature maps was compared with that of radiographic images by radiologists and dental specialists. RESULTS: Significantly improved detection of caries spots was achieved by employing the CLU and FOF texture feature maps. The caries-affected area with sharp margins was well defined using the CLU approach. A pseudo-three-dimensional effect was observed in outlining the demineralization zones inside the cavity with the FOF 5 protocol. In contrast, the LBP and RLM techniques produced less satisfactory results with unsharp edges and less detailed depiction of the lesions. CONCLUSIONS: This study illustrated the applicability of texture feature maps to the recognition of demineralized spots on the tooth surface debilitated by caries and identified the best performing techniques.
OBJECTIVES: Dental caries are caused by tooth demineralization due to bacterial plaque formation. However, the resulting lesions are often discrete and thus barely recognizable in intraoral radiography images. Therefore, more advanced detection techniques are in great demand among dentists and radiographers. This study was performed to evaluate the performance of texture feature maps in the recognition of discrete demineralization related to caries plaque formation. METHODS: Digital intraoral radiology image analysis protocols incorporating first-order features (FOF), co-occurrence matrices, gray tone difference matrices, run-length matrices (RLM), local binary patterns (LBP), and k-means clustering (CLU) were used to transform the digital intraoral radiology images of 10 patients with confirmed caries, which were retrospectively reviewed in a dental clinic. The performance of the resulting texture feature maps was compared with that of radiographic images by radiologists and dental specialists. RESULTS: Significantly improved detection of caries spots was achieved by employing the CLU and FOF texture feature maps. The caries-affected area with sharp margins was well defined using the CLU approach. A pseudo-three-dimensional effect was observed in outlining the demineralization zones inside the cavity with the FOF 5 protocol. In contrast, the LBP and RLM techniques produced less satisfactory results with unsharp edges and less detailed depiction of the lesions. CONCLUSIONS: This study illustrated the applicability of texture feature maps to the recognition of demineralized spots on the tooth surface debilitated by caries and identified the best performing techniques.
Entities:
Keywords:
Caries; Digital intraoral radiography; Image preprocessing; Texture analysis
Authors: Roberta Grassi; Gianna Maria Nardi; Marta Mazur; Roberto Di Giorgio; Livia Ottolenghi; Fabrizio Guerra Journal: Medicina (Kaunas) Date: 2022-04-13 Impact factor: 2.948