Researchers in China are touting a new deep learning-based process that “fuses” multi-modal scans to create a higher quality medical image that can improve clinical diagnosis and patient outcomes.

Known as image fusion, the technique automatically identifies and combines information of scans from different modalities to produce a single high-quality image. “Experimental results indicate that the proposed method achieves state-of-the-art performance in terms of both visual quality and quantitative evaluation metrics,” said author Yi Li, with Qingdao University’s College of Data Science and Software Engineering in Qingdao, China, in a statement.

Li and his colleagues used the technique to fuse MR, CT and SPECT images to build an image training database. The database was then used to fuse medical images in batches, with the newly fused images appearing more natural and with sharper edges and higher resolutions. Additionally, detailed information and features of interest were better preserved to some extent, while key information was clearly contrasted and the virtual shadow was effectively removed.
The end result, according to the researchers, is a single image that contains most of the information in the two multi-modal images of CT and MR or MR and SPECT, thereby making up for the deficiencies of each of these images alone.

“The increase in the amount of information undoubtedly brings changes to the improvement of medical imaging diagnoses. More valuable information can be used to support effective diagnosis. It also brings about possibilities for the study of 'automatic diagnosis' technology and conducts tentative research,” wrote the researchers in their study.

Their findings were published in the June issue of the International Journal of Cognitive Computing in Engineering.

CT Homepage