Study: Computers perform quicker, more accurate readings of MR images
Researchers from Ohio State Medical Center in Columbus have developed a computer program that interprets MR images of the meniscus--a fibrous tissue and cartilage in the knee--to determine if an individual is at risk for developing osteoarthritis, according to research published online Jan. 25 in Osteoarthritis and Cartilage.
Lead author Mark Swanson, a medical student at Ohio State, and colleagues, developed the computer program by creating algorithms based partially on the intensity of the pixels within each component of the images taken of the study participants' knees. The Osteoarthritis Initiative, a national study of the disorder, included the Ohio State Medical Center as one of the four clinical centers used to collect information regarding the treatment of the condition.
The researchers gathered data from 4,796 study participants and randomly selected a cohort of 24 images from this collection, 14 from patients that had been diagnosed with osteoarthritis and 10 patients presenting with no symptoms.
The researchers found that the computer program, which automates measurements of two to two dozen MRI slices of the meniscus, was quicker and just as reliable than human readings of mild to moderate cases of osteoarthritis of the knee.
"We set up a process of elimination for consideration. It says bright pixels are not the meniscus. And we know some areas in the images are bone, ligaments and cartilage, so the algorithms won't let those areas be considered the meniscus," said Swanson. "Once the programming is complete, our algorithms know the anatomy of the knee."
After the program compares the previous slice to the current slice, and re-evaluates each image, the physician must scroll through images manually, find the first slice that includes an image of the meniscus, and place a point within that area of the image. A second point must be placed on the meniscus in the last slice in which that part of the knee anatomy appears.
After the segmentations are complete, the volume, thickness, intensity and any tears in the meniscus can be compared with calculations made from later images. The physician can then determine whether the changes in the meniscus can be associated with symptoms of osteoarthritis, noted the researchers.
The researchers said that when tested against five clinicians specifically trained to manually segment the meniscus within the images, the computer program was equally as skilled as two humans whose interpretations of the same image where compared and exceeded the accuracy of a single person interpreting the same image twice.
In addition, computer segmentation of the images takes between two and four minutes to complete, while manual interpretation can take from seven to 20 minutes, wrote the authors.
If the meniscus can be more precisely measured for changes over time, it may serve as an important biomarker for indicating what individuals are more likely to develop osteoarthritis, concluded Swanson and colleagues.
To further their research, the authors noted plans to advance their computer program to automate the entire reading process and strengthen the program’s ability to measure severely damaged knees.
Lead author Mark Swanson, a medical student at Ohio State, and colleagues, developed the computer program by creating algorithms based partially on the intensity of the pixels within each component of the images taken of the study participants' knees. The Osteoarthritis Initiative, a national study of the disorder, included the Ohio State Medical Center as one of the four clinical centers used to collect information regarding the treatment of the condition.
The researchers gathered data from 4,796 study participants and randomly selected a cohort of 24 images from this collection, 14 from patients that had been diagnosed with osteoarthritis and 10 patients presenting with no symptoms.
The researchers found that the computer program, which automates measurements of two to two dozen MRI slices of the meniscus, was quicker and just as reliable than human readings of mild to moderate cases of osteoarthritis of the knee.
"We set up a process of elimination for consideration. It says bright pixels are not the meniscus. And we know some areas in the images are bone, ligaments and cartilage, so the algorithms won't let those areas be considered the meniscus," said Swanson. "Once the programming is complete, our algorithms know the anatomy of the knee."
After the program compares the previous slice to the current slice, and re-evaluates each image, the physician must scroll through images manually, find the first slice that includes an image of the meniscus, and place a point within that area of the image. A second point must be placed on the meniscus in the last slice in which that part of the knee anatomy appears.
After the segmentations are complete, the volume, thickness, intensity and any tears in the meniscus can be compared with calculations made from later images. The physician can then determine whether the changes in the meniscus can be associated with symptoms of osteoarthritis, noted the researchers.
The researchers said that when tested against five clinicians specifically trained to manually segment the meniscus within the images, the computer program was equally as skilled as two humans whose interpretations of the same image where compared and exceeded the accuracy of a single person interpreting the same image twice.
In addition, computer segmentation of the images takes between two and four minutes to complete, while manual interpretation can take from seven to 20 minutes, wrote the authors.
If the meniscus can be more precisely measured for changes over time, it may serve as an important biomarker for indicating what individuals are more likely to develop osteoarthritis, concluded Swanson and colleagues.
To further their research, the authors noted plans to advance their computer program to automate the entire reading process and strengthen the program’s ability to measure severely damaged knees.