Fig3.

Deep Learning Improves Age Prediction in head CT. Age can be predicted more accurately when using convolutional and structural features on a fully connected network model (A). Most subjects can be predicted within 7.90 years using the combined model (B). Accuracy of prediction does not show characteristic features of aging with our model (C). Note that while the older adult predicted young is 48 years old and technically in the middle aged bin, this was the oldest subject predicted within the young category, so we include it as a proxy.