Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers
Convolutional neural networks (CNNs) have a proven track record in medical image segmentation. Recently, Vision Transformers were introduced and are gaining popularity for many computer vision applications, including object detection, classification, and segmentation. Machine learning algorithms suc...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-10-01
|
| Series: | Tomography |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2379-139X/9/5/151 |
| _version_ | 1827719794405670912 |
|---|---|
| author | Xiaofan Xiong Brian J. Smith Stephen A. Graves Michael M. Graham John M. Buatti Reinhard R. Beichel |
| author_facet | Xiaofan Xiong Brian J. Smith Stephen A. Graves Michael M. Graham John M. Buatti Reinhard R. Beichel |
| author_sort | Xiaofan Xiong |
| collection | DOAJ |
| description | Convolutional neural networks (CNNs) have a proven track record in medical image segmentation. Recently, Vision Transformers were introduced and are gaining popularity for many computer vision applications, including object detection, classification, and segmentation. Machine learning algorithms such as CNNs or Transformers are subject to an inductive bias, which can have a significant impact on the performance of machine learning models. This is especially relevant for medical image segmentation applications where limited training data are available, and a model’s inductive bias should help it to generalize well. In this work, we quantitatively assess the performance of two CNN-based networks (U-Net and U-Net-CBAM) and three popular Transformer-based segmentation network architectures (UNETR, TransBTS, and VT-UNet) in the context of HNC lesion segmentation in volumetric [F-18] fluorodeoxyglucose (FDG) PET scans. For performance assessment, 272 FDG PET-CT scans of a clinical trial (ACRIN 6685) were utilized, which includes a total of 650 lesions (primary: 272 and secondary: 378). The image data used are highly diverse and representative for clinical use. For performance analysis, several error metrics were utilized. The achieved Dice coefficient ranged from 0.833 to 0.809 with the best performance being achieved by CNN-based approaches. U-Net-CBAM, which utilizes spatial and channel attention, showed several advantages for smaller lesions compared to the standard U-Net. Furthermore, our results provide some insight regarding the image features relevant for this specific segmentation application. In addition, results highlight the need to utilize primary as well as secondary lesions to derive clinically relevant segmentation performance estimates avoiding biases. |
| first_indexed | 2024-03-10T20:51:06Z |
| format | Article |
| id | doaj.art-83b20ab732704186bca6974681f43022 |
| institution | Directory Open Access Journal |
| issn | 2379-1381 2379-139X |
| language | English |
| last_indexed | 2024-03-10T20:51:06Z |
| publishDate | 2023-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Tomography |
| spelling | doaj.art-83b20ab732704186bca6974681f430222023-11-19T18:21:05ZengMDPI AGTomography2379-13812379-139X2023-10-01951933194810.3390/tomography9050151Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision TransformersXiaofan Xiong0Brian J. Smith1Stephen A. Graves2Michael M. Graham3John M. Buatti4Reinhard R. Beichel5Department of Biomedical Engineering, The University of Iowa, Iowa City, IA 52242, USADepartment of Biostatistics, The University of Iowa, Iowa City, IA 52242, USADepartment of Radiology, The University of Iowa, Iowa City, IA 52242, USADepartment of Radiology, The University of Iowa, Iowa City, IA 52242, USADepartment of Radiation Oncology, University of Iowa Hospitals and Clinics, Iowa City, IA 52242, USADepartment of Electrical and Computer Engineering, The University of Iowa, Iowa City, IA 52242, USAConvolutional neural networks (CNNs) have a proven track record in medical image segmentation. Recently, Vision Transformers were introduced and are gaining popularity for many computer vision applications, including object detection, classification, and segmentation. Machine learning algorithms such as CNNs or Transformers are subject to an inductive bias, which can have a significant impact on the performance of machine learning models. This is especially relevant for medical image segmentation applications where limited training data are available, and a model’s inductive bias should help it to generalize well. In this work, we quantitatively assess the performance of two CNN-based networks (U-Net and U-Net-CBAM) and three popular Transformer-based segmentation network architectures (UNETR, TransBTS, and VT-UNet) in the context of HNC lesion segmentation in volumetric [F-18] fluorodeoxyglucose (FDG) PET scans. For performance assessment, 272 FDG PET-CT scans of a clinical trial (ACRIN 6685) were utilized, which includes a total of 650 lesions (primary: 272 and secondary: 378). The image data used are highly diverse and representative for clinical use. For performance analysis, several error metrics were utilized. The achieved Dice coefficient ranged from 0.833 to 0.809 with the best performance being achieved by CNN-based approaches. U-Net-CBAM, which utilizes spatial and channel attention, showed several advantages for smaller lesions compared to the standard U-Net. Furthermore, our results provide some insight regarding the image features relevant for this specific segmentation application. In addition, results highlight the need to utilize primary as well as secondary lesions to derive clinically relevant segmentation performance estimates avoiding biases.https://www.mdpi.com/2379-139X/9/5/151head and neck cancersegmentationFDG PETCNNVision Transformer |
| spellingShingle | Xiaofan Xiong Brian J. Smith Stephen A. Graves Michael M. Graham John M. Buatti Reinhard R. Beichel Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers Tomography head and neck cancer segmentation FDG PET CNN Vision Transformer |
| title | Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers |
| title_full | Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers |
| title_fullStr | Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers |
| title_full_unstemmed | Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers |
| title_short | Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers |
| title_sort | head and neck cancer segmentation in fdg pet images performance comparison of convolutional neural networks and vision transformers |
| topic | head and neck cancer segmentation FDG PET CNN Vision Transformer |
| url | https://www.mdpi.com/2379-139X/9/5/151 |
| work_keys_str_mv | AT xiaofanxiong headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers AT brianjsmith headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers AT stephenagraves headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers AT michaelmgraham headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers AT johnmbuatti headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers AT reinhardrbeichel headandneckcancersegmentationinfdgpetimagesperformancecomparisonofconvolutionalneuralnetworksandvisiontransformers |