Radiology: Can the activity of MS lesions be predicted without contrast agents?

 Multiple sclerosis (MS) is the most common non-traumatic demyelinating neurological disease among young people, affecting more than 2.
5 million people worldwide
.

MRI is the most commonly used diagnostic and treatment monitoring method for MS

 Multiple sclerosis (MS) is the most common non-traumatic demyelinating neurological disease among young people, affecting more than 2.

Deep learning (DL) is a subfield of machine learning that uses multiple non-linear processing layers to represent data hierarchically

Recently, a study published in the journal Radiology explored the potential of DL to predict MS enhancement of lesions without using GBCA, and provided technical support for further optimizing the examination and follow-up procedures for MS patients

This study involves a prospective analysis of existing MRI data

This study analyzed 1,008 subjects (mean age 37.

 

Figure 1 Examples of images input to the network (T2, FLAIR and T1 flat scan images)
.

The enhanced T1-weighted image (T1post) shows enhanced areas of true positives (white arrows) and false negatives (black arrows)

Figure 1 Examples of images input to the network (T2, FLAIR and T1 flat scan images)

This study conducted training and testing on MRI data of a large number of patients with multiple sclerosis (MS), which confirmed the feasibility of using deep learning (DL) to predict and enhance lesions on MRI images obtained without gadolinium (GBCAs) The classification accuracy of lesions is 70%

Original source:

Ponnada A Narayana , Ivan Coronado , Sheeba J Sujit ,et al.

A Narayana Ponnada , Ivan Coronado , Sheeba Sujit J .
, Et Al Deep Learning for Predicting Enhancing Lesions in Multiple Sclerosis noncontrast from the MRI .
The DOI: 10.
1148 / radiol.
2019191061 10.
1148 / radiol.
2019191061 in this message