An efficient annotation method for image recognition of dental instruments

Abstract

To prevent needlestick injury and leftover instruments, and to perform efficient dental treatment, it is important to know the instruments required during dental treatment. Therefore, we will obtain a dataset for image recognition of dental treatment instruments, develop a system for detecting dental treatment instruments during treatment by image recognition, and evaluate the performance of the system to establish a method for detecting instruments during treatment. We created an image recognition dataset using 23 types of instruments commonly used in the Department of Restorative Dentistry and Endodontology at Osaka University Dental Hospital and a surgeon’s hands as detection targets. Two types of datasets were created: one annotated with only the characteristic parts of the instruments, and the other annotated with the entire parts of instruments. YOLOv4 and YOLOv7 were used as the image recognition system. The performance of the system was evaluated in terms of two metrics: detection accuracy (DA), which indicates the probability of correctly detecting the number of target instruments in an image, and the average precision (AP). When using YOLOv4, the mean DA and AP were 89.3% and 70.9%, respectively, when the characteristic parts of the instruments were annotated and 85.3% and 59.9%, respectively, when the entire parts of the instruments were annotated. When using YOLOv7, the mean DA and AP were 89.7% and 80.8%, respectively, when the characteristic parts of the instruments were annotated and 84.4% and 63.5%, respectively, when the entire parts of the instruments were annotated. The detection of dental instruments can be performed efficiently by targeting the parts characterizing them.

Image-Based Hand Tools and Accessories Recognition by ResNet50

“…The classification for PPE protection every parts of worker body, which are head with bump caps, eyes with safety goggles, hearing with ear plugs, breathing with respirators, hand with gloves, foot safety toed shoes, skin coats and body with safety vest. Oka et al [4] presented how to recognize 24 kinds of dental instruments in 1425 images with correctly and operate safely by using a convolution neural network. The average precision rates of the system are 89.7%…”

“…Method Tool types Number of Images Precision % Kurnaz [2] Faster RetinaNet 49 2,788 85.00 Oka [4] Yolo V7 24 1,425 80.80 Manici [5] Alex Net 9 4,500 60.00 Stasiak [9] KNN 10 150 85.00 Hernandez [11] Yolo 24 8,000 98.00 This Research ResNet50 110 99,000 99.30…”

Metastatic Lymph Node Detection on Ultrasound Images Using YOLOv7 in Patients with Head and Neck Squamous Cell Carcinoma

“…While this study employed static images as a preliminary step in the development of the DL model for US diagnosis, our ultimate goal is to create a DL model capable of real-time detection of metastatic LNs during live US video examinations. In this study, the YOLO version 7 (YOLOv7) algorithm was used since previous studies have suggested that YOLOv7 provides greater accuracy and requires less computation time [40][41][42][43][44][45]. The salient features of YOLOv7 used in this study are as follows:…”

“…Provides predesigned freebies to facilitate model fine-tuning and simplifies the addition of modules and the creation of new models, which are characterized by higher detection accuracy, speed, and convenience [40][41][42][43][44][45].…”

Automatic Puncture Timing Detection for Multi-Camera Injection Motion Analysis

“…Their results showed that YOLOv7 is faster, but lighter YOLO models can achieve real-time tracking. Oka S. et al [26] developed an image recognition system for detecting dental instruments during treatment to prevent injuries and leftover instruments. YOLOv4 and YOLOv7 were used, with mean detection accuracy ranging from 85.3% to 89.7%.…”