Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Optimal predictive performance of InceptionResNetv2- and NASNetLarge-based models for predicting GTVp and GTVn regression, respectively. ( a ) Heatmap of the AUCs yielded by 25 InceptionResNetv2-based models (all combinations of five machine learning algorithms in rows and five feature selection algorithms in columns) predicting GTVp regression. ( b ) Corresponding heatmap of AUCs for the 25 NASNetLarge-based models predicting GTVn regression.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques: Selection

Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Mean 0.632 + bootstrap areas under the curve (AUCs), sensitivity, and specificity of the deep learning-based radiomics, handcrafted radiomics features, clinical factors, and combined models for predicting primary gross tumor volume (GTVp) regression.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques:

Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Activation maps of the initial CT images reveal salient features used by InceptionResNetv2-based models for prediction of GTVp regression. ( a ) Activation map of the initial CT image from patients with large GTVp regression using InceptionResNetv2 (the CNN yielding the highest predictive accuracy). The map was visualized using the Gradient Weighted Class Activation Mapping method. The boost CT images are shown to indicate the degree of regression. ( b ) Activation map of the initial GTVp images for all patients yielded by InceptionResNetv2.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques: Activation Assay

Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Optimal predictive performance of InceptionResNetv2- and NASNetLarge-based models for predicting GTVp and GTVn regression, respectively. ( a ) Heatmap of the AUCs yielded by 25 InceptionResNetv2-based models (all combinations of five machine learning algorithms in rows and five feature selection algorithms in columns) predicting GTVp regression. ( b ) Corresponding heatmap of AUCs for the 25 NASNetLarge-based models predicting GTVn regression.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques: Selection

Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Mean 0.632 + bootstrap AUCs, sensitivity, and specificity of the deep learning-based radiomics, handcrafted radiomics features, clinical factors, and combined models for predicting nodal gross tumor volume (GTVn) regression.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques:

Journal: Scientific Reports

Article Title: A deep learning-based radiomics approach to predict head and neck tumor regression for adaptive radiotherapy

doi: 10.1038/s41598-022-12170-z

Figure Lengend Snippet: Activation maps of the initial CT images reveal salient features used by NASNetLarge-based models for prediction of GTVn regression. ( a ) Activation map of the initial CT image from a patient with large GTVn regression using NASNetLarge, the CNN yielding the highest prediction accuracy. The boost CT images are shown to illustrate the degree of regression. ( b ) Activation maps for all patients using NASNetLarge.

Article Snippet: We used the deep learning toolbox of MATLAB and downloaded 16 CNNs, SqueezeNet, GoogleNet, Inceptionv3, DenseNet201, MobileNetv2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetv2, ShuffleNet, NASNetMobile, NASNetLarge, DarkNet19, DarkNet53, and AlexNet, all pretrained on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) .

Techniques: Activation Assay