Radiology: Arm artificial intelligence with the thinking of a radiologist!

At this stage, artificial intelligence (AI) has shown great potential in assisting healthcare and medical imaging, among which deep learning is the most influential AI tool
.

Many successful applications of deep learning in neuroradiology are to extract important imaging features such as bleeding on brain CT images

At this stage, artificial intelligence (AI) has shown great potential in assisting healthcare and medical imaging, among which deep learning is the most influential AI tool

In the field of imaging, the differential diagnosis of related disease probabilities rather than just a single best diagnosis is more conducive to the formulation of treatment plans and patient management

Recently, a study published in the journal Radiology used a series of AI tools to develop an AI system that performs computational modeling for radiologists in the three consecutive steps of MRI image analysis, and evaluated its use on cranial MRI.

This retrospective study tested the performance of an artificial intelligence system in the probabilistic diagnosis of 19 common and rare diseases obtained from cranial MRI scans between January 2008 and January 2018

For the accuracy of the first three differential diagnoses, the performance of the AI ​​system (91% correct) is comparable to that of academic neuroradiologists (86% correct; P = .

 

Figure 1 This figure shows examples of FLAIR images for each of the 19 neurological diseases included in the study
.

ADEM = acute diffuse encephalomyelitis, CADASIL = autosomal dominant encephalopathy of cerebral cortex infarction and leukoencephalopathy, CNS = primary central nervous system, HIV = human immunodeficiency virus, MS = multiple sclerosis, NMO = Neuromyelitis optica, PML = progressive multifocal leukoencephalopathy, PRES = reversible posterior encephalopathy syndrome

Figure 1 This figure shows examples of FLAIR images for each of the 19 neurological diseases included in the study

 

The graph in Figure 2 shows the performance comparison of a compound artificial intelligence (AI) system with radiologists of different professional levels
.

A.

The graph in Figure 2 shows the performance comparison of a compound artificial intelligence (AI) system with radiologists of different professional levels

In summary, this research has constructed a compound artificial intelligence (AI) system that can perform computational modeling based on the radiologist's perception and cognitive steps of MRI images of the brain

Original source:

Andreas M Rauschecker , Jeffrey D Rudie , Long Xie , et al.

M Rauschecker andreas , Jeffrey D Rudie , Long Xie , et Al.
Artificial Intelligence Neuroradiologist the System-Level Differential Diagnosis Approaching the Accuracy AT Brain the MRI .
The DOI: 10.
1148 / radiol.
2020190283 10.
1148 / radiol.
2020190283 in this message