MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Deep Learning for 3D Automated Delineation of Primary Gross Tumor Volume for Nasopharyngeal Carcinoma by CT Combining Contrast-Enhanced CT

Deep Learning for 3D Automated Delineation of Primary Gross Tumor Volume for Nasopharyngeal Carcinoma by CT Combining Contrast-Enhanced CT

Z Dai1*, X Wang2, H Jin3, C Cai4, S Zhao5, Y Zhu6, Y Chen7, (1) The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou, Guangdong, CN, (2) The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou, Guangdong, CN, (3) The Second Affiliated Hospital Of Guangzhou University Of Chinese Medicine, Guangzhou, Guangdong, CN, (4) The Second Affiliated Hospital Of Guangzhou University Of Chinese Medicine, Guangzhou, Guangdong, AF, (5) The Second Affiliated Hospital Of Guangzhou University Of Chinese Medicine, Guangzhou, Guangdong, CN, (6) The Second Affiliated Hospital Of Guangzhou University Of Chinese Medicine, Guangzhou, Guangdong, CN, (7) The Second Affiliated Hospital Of Guangzhou University Of Chinese Medicine, Guangzhou, Guangdong, CN

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: present paper is to develop an automated delineation method of nasopharynx gross tumor volume(GTVnx) for nasopharyngeal carcinoma(NPC) in computed tomography(CT) image for radiotherapy applications.

Methods: proposed a modified version of the 3-dimensional(3D) U-Net model with Res-blocks and SE-block for delineation of GTVnx. Besides, an automatic pre-processing method was proposed to crop the 3D region of interest(ROI). Radiotherapy simulation CT images and corresponding manually delineated target of 205 NPC patients diagnosed with stage T1-T4 were used as datasets for training. Automated delineation models were generated based on plain CT(CT) combining contrast-enhanced CT(CE-CT) and CT alone, respectively. We compared the automatic delineation results against the manually delineated contours by radiation oncologists with 5-fold cross-validation to evaluate the performance of the proposed model. We also compared with the framework using 3D convolutional neural network(CNN) and 2-dimensional(2D) deep deconvolutional neural network(DDNN), respectively. Besides, the model generated by one medical group was assessed against the other two separate medical groups. Precision(PR), Sensitivity(SE), Dice Similarity Coefficient(DSC), Average Symmetric Surface Distance(ASSD), and 95% Hausdorff Distance(HD95) were calculated for quantitative evaluation.

Results: overall mean values were 77.85%(PR), 74.93%(SE), 74.21%(DSC), 1.46mm(ASSD), 4.78mm(HD95) with our proposed model based on CT combining CE-CT. The results of metrics were as follows: 94.32%,92.19%,76.15%(PR), 73.62%, 71.51%, 79.87%(SE), 82.70%, 80.54%, 77.97%(DSC), 1.47mm, 1.64mm, 1.70mm(ASSD), 3.74mm, 4.12mm, 4.47mm(HD95) for 3D U-Net, 3DCNN, and 2D DDNN respectively. The proposed method outperforms other automatic methods on the CT images. Automated delineation models based on CT combining CE-CT is superior to that base on CT alone.

Conclusion: modified version 3D U-Net model based on Res-block and SE-block is robust and accurate. It could be useful for the 3D delineation of GTVnx for NPC during the planning of radiotherapy.

Funding Support, Disclosures, and Conflict of Interest: Funding Support:the Guangdong medical scientific research foundation( A2019196)

Keywords

3D, Segmentation

Taxonomy

IM/TH- Image Segmentation Techniques: Modality: CT

Contact Email

Additional Links