TY - JOUR
T1 - COVID-19 AEROSOL SUCTION ROBOT TO ASSIST DENTIST SURGERY BASED ON MOUTH OPENNESS DETECTION USING DEEP LEARNING
AU - Sigit, Riyanto
AU - Yuliyanto, Aditia
AU - Rochmad, Moch
AU - Azhar, Imam Safari
N1 - Publisher Copyright:
© 2023, ICIC International. All rights reserved.
PY - 2023/10
Y1 - 2023/10
N2 - As of November 6, 2022, COVID-19 cases have increased in the Southeast Asia Region by approximately +28%. In dentistry, the COVID-19 infection can be transmitted quickly through aerosol particles. Extra-Oral Suction (EOS) is a mechanism for sucking an aerosol of the patient to counter the transmission of COVID-19; nevertheless the EOS suction nozzle is still manually driven. This allows aerosol particles to escape the mechanism when the patient turns his head or shifts the position of the head. In this case, Visual Servoing (VS) is required; In the realm of image processing, specifically in the context of mouth openness detection, we explore the integration of autonomous control mechanisms to dynamically adjust the orientation of a nozzle. We used a PBVS (Position-Based Visual Servoing) approach integrated with the TensorFlow deep learning models, namely EfficientDet D0, Single Shot Detector (SSD) MobileNet, and SSD ResNet50. Based on the test results of mouth openness detection, EfficientDet has the highest accuracy reaching 100%, SSD MobileNet has an accuracy of 97%, and SSD ResNet50 has an accuracy of 98%. In addition, the system can track the position of the human mouth with a response speed of 135 px/s and an accuracy of 73.1%.
AB - As of November 6, 2022, COVID-19 cases have increased in the Southeast Asia Region by approximately +28%. In dentistry, the COVID-19 infection can be transmitted quickly through aerosol particles. Extra-Oral Suction (EOS) is a mechanism for sucking an aerosol of the patient to counter the transmission of COVID-19; nevertheless the EOS suction nozzle is still manually driven. This allows aerosol particles to escape the mechanism when the patient turns his head or shifts the position of the head. In this case, Visual Servoing (VS) is required; In the realm of image processing, specifically in the context of mouth openness detection, we explore the integration of autonomous control mechanisms to dynamically adjust the orientation of a nozzle. We used a PBVS (Position-Based Visual Servoing) approach integrated with the TensorFlow deep learning models, namely EfficientDet D0, Single Shot Detector (SSD) MobileNet, and SSD ResNet50. Based on the test results of mouth openness detection, EfficientDet has the highest accuracy reaching 100%, SSD MobileNet has an accuracy of 97%, and SSD ResNet50 has an accuracy of 98%. In addition, the system can track the position of the human mouth with a response speed of 135 px/s and an accuracy of 73.1%.
KW - COVID-19
KW - Deep learning
KW - Dentist surgery
KW - Mouth tracking
KW - Visual servoing
UR - http://www.scopus.com/inward/record.url?scp=85173004722&partnerID=8YFLogxK
U2 - 10.24507/ijicic.19.05.1377
DO - 10.24507/ijicic.19.05.1377
M3 - Article
AN - SCOPUS:85173004722
SN - 1349-4198
VL - 19
SP - 1377
EP - 1391
JO - International Journal of Innovative Computing, Information and Control
JF - International Journal of Innovative Computing, Information and Control
IS - 5
ER -