Batch Similarity Based Triplet Loss Assembled into Light-Weighted Convolutional Neural Networks for Medical Image Classification

Discrimination of benign and malignant breast lesions on dynamic contrast-enhanced magnetic resonance imaging using deep learning

PURPOSE

To evaluate the capability of deep transfer learning (DTL) and fine-tuning methods in differentiating malignant from benign lesions in breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI).

METHODS

The diagnostic efficiencies of the VGG19, ResNet50, and DenseNet201 models were tested under the same dataset. The model with the highest performance was selected and modified utilizing three fine-tuning strategies (S1-3). Fifty additional lesions were selected to form the validation set to verify the generalization abilities of these models. The accuracy (Ac) of the different models in the training and test sets, as well as the precision (Pr), recall rate (Rc), F1 score (), and area under the receiver operating characteristic curve (AUC), were primary performance indicators. Finally, the kappa test was used to compare the degree of agreement between the DTL models and pathological diagnosis in differentiating malignant from benign breast lesions.

RESULTS

The Pr, Rc, f1, and AUC of VGG19 (86.0%, 0.81, 0.81, and 0.81, respectively) were higher than those of DenseNet201 (70.0%, 0.61, 0.63, and 0.61, respectively) and ResNet50 (61.0%, 0.59, 0.59, and 0.59). After fine-tuning, the Pr, Rc, f1, and AUC of S1 (87.0%, 0.86, 0.86, and 0.86, respectively) were higher than those of VGG19. Notably, the degree of agreement between S1 and pathological diagnosis in differentiating malignant from benign breast lesions was 0.720 (κ = 0.720), which was higher than that of DenseNet201 (κ = 0.440), VGG19 (κ = 0.640), and ResNet50 (κ = 0.280).

CONCLUSION

The VGG19 model is an effective method for identifying benign and malignant breast lesions on DCE-MRI, and its performance can be further improved via fine-tuning. Overall, our findings insinuate that this technique holds potential clinical application value.

Residual neural network with mixed loss based on batch training technique for identification of EGFR mutation status in lung cancer

Epidermal growth factor receptor (EGFR) is the key to targeted therapy with tyrosine kinase inhibitors in lung cancer. Traditional identification of EGFR mutation status requires biopsy and sequence testing, which may not be suitable for certain groups who cannot perform biopsy. In this paper, using easily accessible and non-invasive CT images, the residual neural network (ResNet) with mixed loss based on batch training technique is proposed for identification of EGFR mutation status in lung cancer. In this model, the ResNet is regarded as the baseline for feature extraction to avoid the gradient disappearance. Besides, a new mixed loss based on the batch similarity and the cross entropy is proposed to guide the network to better learn the model parameters. The proposed mixed loss utilizes the similarity among batch samples to evaluate the distribution of training data, which can reduce the similarity of different classes and the difference of the same classes. In the experiments, VGG16Net, DenseNet, ResNet18, ResNet34 and ResNet50 models with the mixed loss are trained on the public CT dataset with 155 patients including EGFR mutation status from TCIA. The trained networks are employed to the collected preoperative CT dataset with 56 patients from the cooperative hospital for validating the efficiency of the proposed models. Experimental results show that the proposed models are more appropriate and effective on the lung cancer dataset for identifying the EGFR mutation status. In these models, the ResNet34 with mixed loss is optimal (accuracy = 81.58%, AUC = 0.8861, sensitivity = 80.02%, specificity = 82.90%).