Convolutional neural network (CNN)
A neural network built for spatial data like images, where each layer looks at a small neighbourhood of the input and global context only builds up in the deeper layers. CNNs are strong on local patterns but weaker at long-range relationships, which is part of why vision transformers are used as an alternative.
A radiology department compares two wrist fracture detection tools, one built on a convolutional neural network and a newer one built on a vision transformer and marketed as the more advanced design. For a small cortical break, a local pattern of the kind CNNs are built to detect, a well-trained CNN is a strong baseline, and a newer architecture does not by itself predict better performance. Ask both suppliers for sensitivity and specificity on the same set of your own studies, including subtle scaphoid fractures, and compare the results in one table before deciding.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →