Convolutional neural networks: an overview and application in radiology

Abstract

Convolutional neural network (CNN), a class of artificial neural networks that has become dominant in various computer vision tasks, is attracting interest across a variety of domains, including radiology. CNN is designed to automatically and adaptively learn spatial hierarchies of features through backpropagation by using multiple building blocks, such as convolution layers, pooling layers, and fully connected layers. This review article offers a perspective on the basic concepts of CNN and its application to various radiological tasks, and discusses its challenges and future directions in the field of radiology. Two challenges in applying CNN to radiological tasks, small dataset and overfitting, will also be covered in this article, as well as techniques to minimize them. Being familiar with the concepts and advantages, as well as limitations, of CNN is essential to leverage its potential in diagnostic radiology, with the goal of augmenting the performance of radiologists and improving patient care.Key Points • Convolutional neural network is a class of deep learning methods which has become dominant in various computer vision tasks and is attracting interest across a variety of domains, including radiology. • Convolutional neural network is composed of multiple building blocks, such as convolution layers, pooling layers, and fully connected layers, and is designed to automatically and adaptively learn spatial hierarchies of features through a backpropagation algorithm. • Familiarity with the concepts and advantages, as well as limitations, of convolutional neural network is essential to leverage its potential to improve radiologist performance and, eventually, patient care.

Identifying relationships between imaging phenotypes and lung cancer-related mutation status: EGFR and KRAS

“…Also, another work based on deep learning techniques with an interpretable visual output, identified that the regions surrounding the nodule were the most relevant for the classification decision 18,31 . In our opinion, it is crucial to emphasise this characteristic, as it might change the direction and broaden the analysis spectrum of future radiogenomics studies, which until now have been mainly focusing on the nodule or in a region of interest (ROI) around it [32][33][34] . Lung cancer is the result of multiple and complex combinations of morphological, molecular and genetic alterations 35 .…”

Beyond the Randomized Clinical Trial: Innovative Data Science to Close the Pediatric Evidence Gap

“…Deep-learning models have been particularly effective in image analysis, mainly in radiology. 49 A deep-learning model can learn to classify images as healthy or diseased or can notate the areas in the image that correspond to organs or other anatomic structures. For example, a deep-learning model was able to identify the segmentation of white matter, gray matter, and cerebrospinal fluid in the brains of babies.…”

Measuring Particle Size Distributions in Multiphase Flows Using a Convolutional Neural Network

“…In every network layer, there can be a number of different convolution steps resulting in different feature maps. To the output of the convolution, a nonlinear activation function is applied, here, the most common function is the rectified linear unit (ReLU) operation, which is : …”

“…This can be overcome by padding the edges with an extra layer. [5] The kernels are the learnable parameter in this operation and are adapted in the training of the network. In every network layer, there can be a number of different convolution steps resulting in different feature maps.…”

“…1) and returns the maximum value of this patch as output. [5] An advantage of U-Net is that it can produce a local segmentation. The output of the network is a segmentation mask, which can be used to analyze images considering the diameter of the segmented object.…”