Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 59, Issue 8, 0800002(2022)

Application of Convolution Neural Network in Diagnosis of Thyroid Nodules

Xuanqi Wang1, Feng Yang1, Bin Cao2, Jing Liu1, Dejian Wei1, and Hui Cao1、*
Author Affiliations
  • 1College of Intelligence and Information Engineering, Shandong University of Traditional Chinese Medicine, Jinan , Shandong 250355, China
  • 2Shandong Provincial Hospital of Traditional Chinese Medicine, Jinan , Shandong 250000, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (9)
    Equations (0)
    References (71)
    Cited By (2)
    Tools
    Paper Information
    Topics
    Figures & Tables(9)
    Schematic of thyroid nodule detected by Faster RCNN

    Fig. 1. Schematic of thyroid nodule detected by Faster RCNN

    Download full size
    Results of different thyroid nodule detection algorithms. (a) Xie et al[13]; (b) Fang et al[17]; (c) Yu et al[19]; (d) Wang et al[21]; (e) Li et al[23]; (f) Abdolali et al[24]

    Fig. 2. Results of different thyroid nodule detection algorithms. (a) Xie et al[13]; (b) Fang et al[17]; (c) Yu et al[19]; (d) Wang et al[21]; (e) Li et al[23]; (f) Abdolali et al[24]

    Download full size
    Schematic of thyroid nodule segmentated by U-Net

    Fig. 3. Schematic of thyroid nodule segmentated by U-Net

    Download full size
    Caliper marks of thyroid nodules in ultrasound images[33]. (a) Transverse view of cross caliper; (b) longitudinal view of single caliper

    Fig. 4. Caliper marks of thyroid nodules in ultrasound images[33]. (a) Transverse view of cross caliper; (b) longitudinal view of single caliper

    Download full size
    Results of some thyroid nodule segmentation algorithms. (a) Ying et al[30]; (b) Buda et al[33]; (c) Kumar et al[34]

    Fig. 5. Results of some thyroid nodule segmentation algorithms. (a) Ying et al[30]; (b) Buda et al[33]; (c) Kumar et al[34]

    Download full size

    • Table 1. Performance comparison of different thyroid nodule detection algorithms

      View table

      Table 1. Performance comparison of different thyroid nodule detection algorithms

      ReferenceMethodAverage Precision /%Recall /%Note
      Xie et al［13］SSD88.0890.08General nodule
      Fang et al［17］Faster RCNN92.7989.24General nodule
      Yu et al［19］Faster RCNN97.2093.70General nodule
      Wang et al［21］Faster RCNN88.80Papillary thyroid carcinoma
      Li et al［23］Faster RCNN93.50Papillary thyroid carcinoma
      Abdolali et al［24］Mask RCNN84.0079.00General nodule

    • Table 2. Performance comparison of different thyroid nodule segmentation algorithms

      View table

      Table 2. Performance comparison of different thyroid nodule segmentation algorithms

      ReferenceMethodDice similarity coefficient
      Ying et al［30］FCN+U-Net0.9304
      Zhou et al［31］Mark-Guided U-Net0.9482
      Chu et al［32］Mark-Guided U-Net0.9576
      Buda et al［33］Faster RCNN+U-Net0.9310
      Kumar et al［34］MPCNN0.7600

    • Table 3. Performance comparison of thyroid nodule benign and malignant recognition models mentioned in this paper

      View table

      Table 3. Performance comparison of thyroid nodule benign and malignant recognition models mentioned in this paper

      ReferenceMethodmAP /%Accuracy /%Sensitivity /%Specificity /%
      Moussa et al［44］ResNet-50\97.3380.6964.17
      Song et al［45］Inception-V3\\95.2/94.061.8/56.0
      Guan et al［47］Inception-V3\90.593.387.4
      Zhu et al［48］ResNet-18\93.7593.9692.68
      Zhang et al［49］GoogLeNet\96.04\\
      Wang et al［50］Inception-ResNet-v2（with attention-based feature aggregation network）\87.3284.22\
      Wang et al［25］YOLOv2（backbone： ResNet-50）85.9290.3190.589.91
      Ma et al［26］YOLOv3（backbone： DMRF-CNN）95.2395.2497.39\
      Ma et al［54］CNN+CNN\83.0282.4184.96
      Zheng et al［55］Xception+LSTM+CNN\94.30\\
      Li et al［58］ResNet-50+Darknet-19\

      89.80

      86.50

      85.70

      93.40

      84.70

      84.30

      86.1

      87.8

      86.9

      Liu et al［59］VGG-16+EEGNet\89.688.591.0
      Liang et al［61］ResNet-50+AlexNet+VGG-16\96.094.197.7
      Song et al［27］SSD+AlexNet98.292.194.196.2
      Liu et al［28］Multi-scale detection network （backbone：ResNet-50）+Multi-branch classification network （based on ZFNet）94.797.198.295.1

    • Table 4. Comparison of calcification recognition accuracy between traditional algorithm and CNN

      View table

      Table 4. Comparison of calcification recognition accuracy between traditional algorithm and CNN

      ReferenceMethodAccuracy /%
      Chen et al［64］Threshold segmentation algorithm based on brightness feature62.59
      Choi et al［65］Local Otsu threshold algorithm67.85
      Han et al［66］Maximum extremum stable region algorithm69.72
      Zuo et al［67］AlexNet（CNN）86.06
      Zhang et al［68］CS-AGnet（CNN）92.10
    Tools

    Get Citation

    Copy Citation Text

    Xuanqi Wang, Feng Yang, Bin Cao, Jing Liu, Dejian Wei, Hui Cao. Application of Convolution Neural Network in Diagnosis of Thyroid Nodules[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0800002

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Reviews

    Received: Mar. 16, 2021

    Accepted: Apr. 22, 2021

    Published Online: Apr. 18, 2022

    The Author Email: Cao Hui (caohui63@163.com)

    DOI:10.3788/LOP202259.0800002

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology