Evaluation of deep learning-based multiparametric MRI oropharyngeal primary tumor auto-segmentation and investigation of input channel effects: Results from a prospective imaging registry

Abstract

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer

“…We used a multi-stage deep learning convolutional neural network approach for SM segmentation. Our approach was based on the UNet architecture with residual connections (ResUNet) included in the MONAI software package, as we have described in previous publications ( 33 , 35 ). In the first stage of our approach ( Figure 1C ), a 3D ResUNet model auto-segmented the C3 vertebra section (33 mm), which was then followed by auto-selection of the middle slice of the section.…”

Auto-detection and segmentation of involved lymph nodes in HPV-associated oropharyngeal cancer using a convolutional deep learning neural network

“…A DL-CNN was developed based on a 3-dimensional (3D) residual U-Net architecture included in the Medical Open Network for Artificial Intelligence (MONAI) software package [11] . This architecture has been utilized successfully in previous OPC tumor auto-segmentation studies [12] , [13] . The network consisted of 4 convolution blocks in the encoding and decoding branches with a bottleneck convolution block separating these two branches ( Fig.…”

“…Observers were asked to rate the lymph node segmentations on a Likert scale for clinical acceptability (1 = clinically acceptable, highly accurate; 2 = clinically acceptable, errors not significant; 3 = requires correction, minor errors; 4 = requires correction, large errors). Using a modified Turing test, observers were then asked to determine whether the segmentation was generated by a human or a computer, [12] , [16] . Lastly, observers were asked to rate their confidence in this determination using a Likert scale (1 = very confident; 2 = somewhat confident; 3 = somewhat unconfident; 4 = very unconfident).…”

Deep Learning Auto-Segmentation of Cervical Neck Skeletal Muscle for Sarcopenia Analysis Using Pre-Therapy CT in Patients with Head and Neck Cancer

“…We used a multi-stage deep learning convolutional neural network approach for SM segmentation. Our approach was based on the UNet architecture with residual connections (ResUNet) included in the MONAI software package, as we have described in previous publications [31,33]. In the first stage of our approach (Figure 1C), a 3D ResUNet model autosegmented the C3 vertebra section (33 mm), which was then followed by auto-selection of the middle slice of the section.…”