RESUMEN
Histopathologic whole-slide images (WSI) are generally considered the gold standard for cancer diagnosis and prognosis. Survival prediction based on WSI has recently attracted substantial attention. Nevertheless, it remains a central challenge owing to the inherent difficulties of predicting patient prognosis and effectively extracting informative survival-specific representations from WSI with highly compounded gigapixels. In this study, we present a fully automated cellular-level dual global fusion pipeline for survival prediction. Specifically, the proposed method first describes the composition of different cell populations on WSI. Then, it generates dimension-reduced WSI-embedded maps, allowing for efficient investigation of the tumor microenvironment. In addition, we introduce a novel dual global fusion network to incorporate global and inter-patch features of cell distribution, which enables the sufficient fusion of different types and locations of cells. We further validate the proposed pipeline using The Cancer Genome Atlas lung adenocarcinoma dataset. Our model achieves a C-index of 0.675 (±0.05) in the five-fold cross-validation setting and surpasses comparable methods. Further, we extensively analyze embedded map features and survival probabilities. These experimental results manifest the potential of our proposed pipeline for applications using WSI in lung adenocarcinoma and other malignancies.
RESUMEN
Purpose: To train a convolutional neural network (CNN) model from scratch to automatically detect tuberculosis (TB) from chest X-ray (CXR) images and compare its performance with transfer learning based technique of different pre-trained CNNs. Material and methods: We used two publicly available datasets of postero-anterior chest radiographs, which are from Montgomery County, Maryland, and Shenzhen, China. A CNN (ConvNet) from scratch was trained to automatically detect TB on chest radiographs. Also, a CNN-based transfer learning approach using five different pre-trained models, including Inception_v3, Xception, ResNet50, VGG19, and VGG16 was utilized for classifying TB and normal cases from CXR images. The performance of models for testing datasets was evaluated using five performances metrics, including accuracy, sensitivity/recall, precision, area under curve (AUC), and F1-score. Results: All proposed models provided an acceptable accuracy for two-class classification. Our proposed CNN architecture (i.e., ConvNet) achieved 88.0% precision, 87.0% sensitivity, 87.0% F1-score, 87.0% accuracy, and AUC of 87.0%, which was slightly less than the pre-trained models. Among all models, Exception, ResNet50, and VGG16 provided the highest classification performance of automated TB classification with precision, sensitivity, F1-score, and AUC of 91.0%, and 90.0% accuracy. Conclusions: Our study presents a transfer learning approach with deep CNNs to automatically classify TB and normal cases from the chest radiographs. The classification accuracy, precision, sensitivity, and F1-score for the detection of TB were found to be more than 87.0% for all models used in the study. Exception, ResNet50, and VGG16 models outperformed other deep CNN models for the datasets with image augmentation methods.
