Optics and Precision Engineering, Volume. 32, Issue 9, 1420(2024)

Segmentation network for metastatic lymph nodes of head and neck tumors

Tao ZHOU1...2, Daozong SHI1,2,*, Jiawen XUE3, Caiyue PENG1,2, Pei DANG1,2 and Zhongwei ZHOU3 |Show fewer author(s)
Author Affiliations
  • 1College of Computer Science and Engineering, North Minzu University, Yinchuan75002, China
  • 2Key Laboratory of Image and Graphics Intelligent Processing of State Ethnic Affairs Commission, North Minzu University, Yinchuan75001, China
  • 3College of Oral Cavity, Ningxia Medical University, Yinchuan750004, China
  • show less
    Figures & Tables(11)
    Overall structure of LNMS Net model
    Cross layer and cross field attention module
    Multi-scale feature fusion module
    Enhanced attention prediction head module
    Dataset presentation
    Visualization results of ablation experiments
    Visualization results of validation experiments
    • Table 1. Design of ablation experiments

      View table
      View in Article

      Table 1. Design of ablation experiments

      消融

      实验

      ResNet101FPNHeadC2AMMFFMEAPM
      ×××
      ××
      ×
    • Table 2. Evaluation indicators and related formulas

      View table
      View in Article

      Table 2. Evaluation indicators and related formulas

      指标作 用公 式
      IoUIoU用于测量标注框和预测框之间的相关度,数值越大,表示相关度越高,模型性能越好。IOU=TP/(TP+FP+FN)
      AP平均精度(Average Precision,AP)是正确识别的目标个数占总识别的目标个数的百分数,用于衡量在每个类别上模型检测器的性能优劣。其中,AP50表示IoU阈值为0.5时的AP值。数值越大,模型性能越好。AP=1t1thtTP(t)TP(t)+FP(t)
      AR平均召回率(Average Recall,AR)是正确识别的目标个数占测试集中识别的目标个数的百分数。其中,AR50表示IoU阈值为0.5时的AR值。数值越大,模型性能越好。AR=1t1thtTP(t)TP(t)+FN(t)
      mAP均值平均精度(Mean Average Precision ,mAP)是衡量检测精度的指标。其中,mAP50表示IoU阈值为0.5时的mAP值。数值越大,该模型性能越好。mAP=1CC1t1thTP(t)TP(t)+FP(t)
    • Table 3. Results of ablation experiment (IoU=0.50)

      View table
      View in Article

      Table 3. Results of ablation experiment (IoU=0.50)

      实验APdetAPsegARdetARsegmAPdetmAPseg
      70.3770.5860.1259.7870.2270.43
      71.9671.7761.2361.0171.6771.58
      73.1272.8662.4061.5773.0272.66
      74.8874.1263.1162.2874.6474.04
    • Table 4. Comparative experimental results (IoU=0.50)

      View table
      View in Article

      Table 4. Comparative experimental results (IoU=0.50)

      模 型APdetAPsegARdetARsegmAPdetmAPseg
      Yolact(Resnet50)68.8969.9458.7757.6869.0370.13
      Yolact(Resnet101)70.3770.5860.1259.7870.2270.43
      Yolact++(Resnet50)72.8672.4561.7861.3372.9772.56
      Yolact++(Resnet101)73.9173.3862.8862.5674.0273.88
      MaskRcnn(Resnet50)72.7772.5661.8661.3572.6872.47
      MaskRcnn(Resnet101)74.0573.7862.6862.0273.8973.65
      LNMS Net74.8874.1263.1162.2874.6474.04
    Tools

    Get Citation

    Copy Citation Text

    Tao ZHOU, Daozong SHI, Jiawen XUE, Caiyue PENG, Pei DANG, Zhongwei ZHOU. Segmentation network for metastatic lymph nodes of head and neck tumors[J]. Optics and Precision Engineering, 2024, 32(9): 1420

    Download Citation

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

    Category:

    Received: Nov. 10, 2023

    Accepted: --

    Published Online: Jun. 2, 2024

    The Author Email: SHI Daozong (shidaozong167@163.com)

    DOI:10.37188/OPE.20243209.1420

    Topics