CHICAGO, Nov. 26—Chest radiography is one of the most commonly performed diagnostic imaging examinations. A technique developed by a group of physicists from the department of radiology at the University of Chicago and Kumamoto University in Japan could further extend the diagnostic potential of digital chest radiography through the application of virtual dual-energy capabilities, according to a physics poster presentation at the 93rd annual meeting of the Radiological Society of North America (RSNA).
Dual-energy imaging exploits the differing physical properties of soft-tissue and bony structures that affect the attenuation of x-ray photons at different x-ray energies. The application of the technique commonly uses dedicated hardware to capture a low-energy image and a high-energy image during a single examination. This results in the construction of a pair of energy subtraction images, allowing either bone or soft tissue to be masked for interpretation.
A technique utilizing a massive training artificial neural network (MTANN) can perform both enhancement of a specific opacity and suppression of other opacities in medical images.
With virtual dual-energy radiography (VDER), rib and soft-tissue components of chest radiographs can be separated by software (without specialized equipment), and soft-tissue and bone images can be produced.
“VDER can provide improved conspicuity of nodules compared to standard chest images and improved enhancement of pathologic changes in temporally subtracted images,” the researchers wrote.
They noted that lung nodules in chest x-rays are often partially obscured by overlying bone such as the ribs or clavicle. Although dual-energy radiography can address this issue, most facilities do not deploy the technology due to the specialized hardware required for its utilization.
“The major advantages of simulated dual-energy radiography compared to conventional dual energy are:
- No additional radiation dose to patients is required
- No specialized equipment for generating dual-energy images is required,” the authors wrote.
consists of a multi-resolution MTANN, which is a supervised non-linear filter consisting of a linear output artificial neural network model capable of operating on image data directly. The developer’s scheme for the MTANN tool has two phases, a training phase and an application phase.
In the training phase, the MTANN is trained with input from chest x-rays and corresponding teaching bone images obtained with a dual-energy radiography system.
“In the application phase, the trained MTANN is applied to a standard chest x-ray to provide a bone-image-like image where the ribs are extracted,” the authors wrote. “The bone-image-like images can then be subtracted from the original image to create a soft-tissue-like image where the ribs are suppressed.”
The tool allows for dual-energy images to be obtained from a standard chest x-ray without additional radiation dose to the patient or the use of any specialized equipment.
The MTANN VDER technique also can be used to improve chest x-ray computer-assisted detection (CAD) technology. One of the limitations of chest x-ray CAD is a high rate of false positives, primarily due to the ribs overlying soft tissue of standard images.
With the VDER tool, bones can be suppressed, and the false positive rate can be decreased. The team found that its application brought the CAD false positive rate down to 1 per image from a prior 4 per image on standard chest x-rays while maintaining sensitivity over 82 percent.
Another technique explored by the physicists was temporal subtraction, which involves subtraction between current and prior chest x-rays of the same patient to better visualize pathologic changes. One of the limitations of temporal subtraction is that misregistration of the images due to rib artifacts can obscure abnormalities. Utilizing the VDER application, the team was able to suppress the ribs and allow for better temporal subtraction image comparison than with conventional chest x-rays.