A retrospective study of deep learning generalization across two centers and multiple models of X-ray devices using COVID-19 chest-X rays.

Generalization of deep learning (DL) algorithms is critical for the secure implementation of computer-aided diagnosis systems in clinical practice. However, broad generalization remains to be a challenge in machine learning. This research aims to identify and study potential factors that can affect the internal validation and generalization of DL networks, namely the institution where the images come from, the image processing applied by the X-ray device, and the type of response function of the X-ray device. For these purposes, a pre-trained convolutional neural network (CNN) (VGG16) was trained three times for classifying COVID-19 and control chest radiographs with the same hyperparameters, but using different combinations of data acquired in two institutions by three different X-ray device manufacturers. Regarding internal validation, the addition of images from an external institution to the training set did not modify the algorithm's internal performance, however, the inclusion of images acquired by a device from a different manufacturer decreased the performance up to 8% (p < 0.05). In contrast, generalization across institutions and X-ray devices with the same type of response function was achieved. Nonetheless, generalization was not observed across devices with different types of response function. This factor was the key impediment to achieving broad generalization in our research, followed by the device's image-processing and the inter-institutional differences, which both reduced generalization performance to 18.9% (p < 0.05), and 9.8% (p < 0.05), respectively. Finally, clustering analysis with features extracted by the CNN was performed, revealing a substantial dependence of feature values extracted by the pre-trained CNN on the X-ray device which acquired the images.

Novel deep learning method for coronary artery tortuosity detection through coronary angiography.

Coronary artery tortuosity is usually an undetected condition in patients undergoing coronary angiography. This condition requires a longer examination by the specialist to be detected. Yet, detailed knowledge of the morphology of coronary arteries is essential for planning any interventional treatment, such as stenting. We aimed to analyze coronary artery tortuosity in coronary angiography with artificial intelligence techniques to develop an algorithm capable of automatically detecting this condition in patients. This work uses deep learning techniques, in particular, convolutional neural networks, to classify patients into tortuous or non-tortuous based on their coronary angiography. The developed model was trained both on left (Spider) and right (45°/0°) coronary angiographies following a fivefold cross-validation procedure. A total of 658 coronary angiographies were included. Experimental results demonstrated satisfactory performance of our image-based tortuosity detection system, with a test accuracy of (87 ± 6)%. The deep learning model had a mean area under the curve of 0.96 ± 0.03 over the test sets. The sensitivity, specificity, positive predictive values, and negative predictive values of the model for detecting coronary artery tortuosity were (87 ± 10)%, (88 ± 10)%, (89 ± 8)%, and (88 ± 9)%, respectively. Deep learning convolutional neural networks were found to have comparable sensitivity and specificity with independent experts' radiological visual examination for detecting coronary artery tortuosity for a conservative threshold of 0.5. These findings have promising applications in the field of cardiology and medical imaging.

Artificial intelligence to estimate wine volume from single-view images.

In this paper, we present a method to determine the volume of wine in different types of glass liquid containers from a single-view image. The proposed model predicts red wine volume from a photograph of the glass containing the wine. Experimental results demonstrated satisfactory performance of our image-based wine measurement system, with a Mean Absolute Error lower than 10 mL . To train and evaluate our system, we introduced the WineGut_BrainUp dataset, a new dataset of glasses of wine that contains 24305 laboratory images, including a wide range of containers, volumes of wine, backgrounds, object distances, angles and lightning, with or without calibration object. The proposed methodology is a suitable analytical tool for automate measurement of red wine volume. Indeed, it has potential real life applications in diet monitoring and wine consumption studies.