Research on thyroid nodule detection model based on deep learning algorithm Mask R-CNN

Corresponding author: CAO Xin-zhi, to_cxz@163.com

DOI: 10.12201/bmr.202411.00085

•  

  Abstract: Purpose/Significance Establish an object detection model through the target mask segmentation algorithm based on Mask Region-based Convolutional Neural Network (Mask R-CNN) to intelligently identify the nodule location in thyroid ultrasound images and provide a reference for the decision-making of ultrasound doctors. Method/Process Collect 165 0 ultrasound nodule images, use the labelme tool to label the nodule locations. Replace the backbone network of Mask R-CNN with MobileNetV3, ResNet50, ResNet101, and ResNet152, and introduce Feature Pyramid Network (FPN) and Region of Interest Align (ROI Align). Train the model using transfer learning training strategy and compare the object detection performance under different networks. Result/Conclusion The backbone network trained using ResNet101 has an average accuracy of 86.8%, an average recall rate of 95.3%, and an average F1 score of 90.6%, which is superior to other backbone networks and can more accurately detect thyroid nodules, it has certain clinical application value.

  Key words: thyroid nodule; Mask R-CNN; Object detection; Neural network

  Submit time: 29 November 2024

  Copyright: The copyright holder for this preprint is the author/funder, who has granted biomedRxiv a license to display the preprint in perpetuity.
• html

• 图表

• liuqingjin, wangrui, miaoyuanqing. Intelligent Detection of Silent Myocardial Ischemia Dynamic Electrocardiogram Based on Deep Learning. 2021. doi: 10.12201/bmr.202111.00009

  李威, Geng Zhong li, Angshaer·Sulaiman. Innovation and evolution of neural monitoring technology in endoscopic thyroid cancer surgery. 2024. doi: 10.12201/bmr.202411.00051

  Zhang Chunyang, Hu Hongwei. Research on the prediction mechanism of working ability of elderly nursing staff based on neural network: Perspective of professionalization. 2021. doi: 10.12201/bmr.202104.00022

  xiayu, yaoyuan, suyue, wushaohua. Study on Quality of Life of Medical Staff and Its Influencing Factors During Sudden Public Health Emergencies Based on Decision Tree and Neural Network Model. 2024. doi: 10.12201/bmr.202412.00006

  Gu Yao-wen, Li Jiao. Progress of Mining Electronic Health Records based on Unsupervised Deep Learning Methods. 2021. doi: 10.12201/bmr.202104.00013

  Zou Yang, Li Jia-ying. Experience of lung nodule treatment by Tongluo-dispersing method. 2024. doi: 10.12201/bmr.202411.00055

  WU Shengnan, WU Jiahui, DONG Jizong, JIANG Huanyu, WANG Luqi, WANG Xinyao. Research on predicting medical entity relationships based on bipartite network representation learning. 2024. doi: 10.12201/bmr.202407.00041

  cuitao, KOU De-shuang. The Research of Tongue Features Base on Deep Learning. 2024. doi: 10.12201/bmr.202404.00020

  HU Haiyang, ZHAO Congpu, Ma Lian, JIANG Huizhen, ZHANG Jing, ZHU Weiguo. Attention Mechanism And Dilated Convolution Neural Networks for Named Entity Recognition. 2021. doi: 10.12201/bmr.202102.00004

  duxuejie, gehui. Study on the design of prediction and early warning model of hand, foot and mouth disease based on BP neural network.. 2021. doi: 10.12201/bmr.202102.00002

• ID	Submit time	Number	Download
  1	2024-08-23	bmr.202411.00085V1	Download

• Public  Anonymous  To author only Submit Comments