The methods of AI remain relatively uncommon in clinical practice for at least two reasons.

First,machine-learing algorithms may not perform well when applied to new data

Second,even successful AI technologies may ultimately have little impact on patient care unless data scientists and physicians collaborate to work out how best to integrate them into clinical practice in real-world setting to improve patient outcomes

The author introduced two reports of new AI technology for CAD of acute neurological events and  retinal disease that succeed by addressing both of the two challenges

The first report use a ANN to decide whether a CT image-typically acquired after a patient presents to an emergency department- contains a critical finding,such as a stroke or hemorrhage.The model of ANN in this report using a large(n=37236) dataset of CT images.they also test 180 images ,for which labels were obtained through manual review of patient medical records by a physicain.The sensitivity of their algorithm was on par wth that of three physicians,albeit with lower specificity (0.48 versus 0.85).They also performed a double-blinded prospective trial in a simulated clinical environment to evaluate whether their model could function effectively as a triage system.It performs well.In this report they conclude their findings suggest that machine learning-based triage system can reduce the time to treatment for urgent case of acute neurologic illness,thereby,imporving patient outcomes

The second report show how AI can used for CAD of retinal disease.The author construct a two-stage ANN to identify retinal pathologies in OCT scans.One key to their success was to design a two-stage algorithm that first accout for technical variations in the image produced by different devices and then diagnosea various retinal diseases.Remarkably,referral accuracies were on par or exceeded those from a group of eight retinal specialists and optometrists,even when these human experts also considered clinical notes and other forms of retinal imaging data.

The conclution of this paper is defining and understanding failure modes will critical as AI technologies become more widely used in clinical settings,and it's also important to bear in mind that human physicians do not need ML method to interpret medical images