Release date: 2017-06-26
Scanning the brain with magnetic resonance imaging (MRI) produces a number of 2-D "slices" that can be synthesized to build a 3-D presentation of the brain. Stroke patients often scan their brains in the hospital so that doctors can identify and identify damaged locations and areas. It usually takes 30 minutes to perform a high-resolution scan of the brain, but in the case of a stroke, this time is too long. Therefore, when a hospital receives a stroke patient, a rapid brain scan is performed. The interval between such fast scanning slices is 7 mm, while in high-end scientific research, slice scanning is usually performed at intervals of 1 mm. Doctors will analyze the effects of stroke on these scans, but often the resolution of such rapid clinical scans is too low, making many computer algorithms for assisted analysis difficult.
Artificial intelligence makes the original image with very low resolution clear (Source: MIT)
“These images are unique because they are obtained in clinical practice when patients are admitted to hospital due to a stroke,†said Polina Golland, professor of electrical engineering and computer science at MIT. “This kind of research is difficult for you to plan.â€
These hospital-based stroke scans are a wealth of data. To help scientists make better use of these scans, the research team from MIT, in collaboration with doctors and other institutions at the Massachusetts General Hospital, developed a way to improve the quality of these scans. Methods so that these clinical data can be used for large-scale stroke studies. Through these scans, researchers can investigate how genetic factors influence stroke survival and how patients respond to different treatment options. They can also use this route to study other diseases, such as Alzheimer's disease.
Fill in the data
For clinical scans of stroke patients, imaging is performed quickly due to the time limit of the scan, and the "slices" of the scan are very sparse, meaning that the sliced ​​images have a 5-7 mm gap. (On-chip resolution is 1mm)
Polina Golland, Professor of Electrical Engineering and Computer Science at MIT (Source: MIT)
For scientific research, researchers often need to obtain higher resolution imaging with a gap of only 1 mm between slices, which requires a longer scan. Scientists have developed specialized computer algorithms to analyze these images, but these algorithms are not very suitable for low-quality hospital scan imaging.
MIT researchers work with researchers from Massachusetts General Hospital and other hospitals who are very interested in how to use these large numbers of patient scan images. These rich clinical resources allow them to learn more than those small, high-quality scan studies.
Together they developed a new approach that basically fills in the missing data in each patient scan, by taking information from the entire scan set and using it to recreate the missing anatomical features in other scans.
Golland said: "The key concept is to generate anatomically reasonable images, using algorithms to make them look like those research scans, and exactly the same as the clinical images obtained. Once you do this, you can apply the most advanced developments. Algorithm for obtaining beautiful research imaging and running the same analysis as a scientific image and getting the results."
A major limitation of ordinary MRI images is resolution (Source: MIT)
Once these images of scientific quality are generated, the researchers can run a set of algorithms designed to help analyze the anatomical features, including the alignment of the slices and the process called skull stripping—in addition to removing structural imaging from the brain.
Throughout the process, the algorithm keeps track of which pixels are coming from the original scan and which pixels are populated later for later analysis. For example, an analysis that measures the extent of brain damage can only be performed on the information of the original scan.
"In a sense, this is a scaffold that allows us to put an image into a collection as if it were a high-resolution image and then only measure the pixels we have information on," Golland said.
Higher quality
The MIT team developed the technology to enhance low-quality images, and now they plan to apply it to about 4,000 stroke imaging scans, including 12 hospitals.
"Understanding the spatial patterns of damage to the white matter helps us understand in more detail how the disease interacts with cognitive ability, and the ability to recover from stroke, etc.," Golland noted.
Artificial intelligence will change our treatment of disease (Source: SiliconANGLE)
The researchers also hope to apply this technology to scans of patients with other brain diseases.
“It opens up a lot of interesting directions,†Golland said. “The images obtained in everyday medical practice can give anatomical insight because we are improving the quality to the extent that the algorithm can analyze.â€
Reference material
[1] New technique makes brain scans better
[2] Artificial Intelligence Helps Improve MRI Imaging of Strokes
Source: Health New Vision (Micro Signal HealthHorizon)
Horizontal Arm Cmm,Nebula Horizontal Arm Cmm,Saturn Horizontal Arm Cmm,Horizontal Cmm With Multi-Functional
Automated Precision Inc. , https://www.apiasean.com