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.

BACKGROUND Desmoid tumors (also referred to as aggressive fibromatosis) are connective tissue neoplasms that can arise in any anatomical location and infiltrate the mesentery, neurovascular structures, and visceral organs. There is no standard of care. METHODS In this double-blind, phase 3 trial, we randomly assigned 87 patients with progressive, symptomatic, or recurrent desmoid tumors to receive either sorafenib (400-mg tablet once daily) or matching placebo. Crossover to the sorafenib group was permitted for patients in the placebo group who had disease progression. The primary end point was investigator-assessed progression-free survival; rates of objective response and adverse events were also evaluated. RESULTS With a median follow-up of 27.2 months, the 2-year progression-free survival rate was 81% (95% confidence interval [CI], 69 to 96) in the sorafenib group and 36% (95% CI, 22 to 57) in the placebo group (hazard ratio for progression or death, 0.13; 95% CI, 0.05 to 0.31; P<0.001). Before crossover, the objective response rate was 33% (95% CI, 20 to 48) in the sorafenib group and 20% (95% CI, 8 to 38) in the placebo group. The median time to an objective response among patients who had a response was 9.6 months (interquartile range, 6.6 to 16.7) in the sorafenib group and 13.3 months (interquartile range, 11.2 to 31.1) in the placebo group. The objective responses are ongoing. Among patients who received sorafenib, the most frequently reported adverse events were grade 1 or 2 events of rash (73%), fatigue (67%), hypertension (55%), and diarrhea (51%). CONCLUSIONS Among patients with progressive, refractory, or symptomatic desmoid tumors, sorafenib significantly prolonged progression-free survival and induced durable responses. (Funded by the National Cancer Institute and others; ClinicalTrials.gov number, NCT02066181.)

No abstract