Curing Alzheimer’s, or at least slowing it, has been a goal of countless researchers, and remains a top priority. In a new review, Chinese physicians describe an imaging approach that may take the specialty one step closer to its objective.
Magnetic resonance texture analysis is part of a growing field that uses math to detect changes in imaging signals invisible in image pixels, researchers wrote Feb. 11 in Academic Radiology. The technique could be used to develop neuroimaging biomarkers, aid clinicians' diagnosis of Alzheimer’s and further research on the disease.
“We are entering a new era of combined imaging and clinical evaluation that will increasingly depend on neuroimaging markers in disease detection and diagnosis,” Jia-Hui-Cai, with the First Affiliated Hospital of University of South China, and colleagues wrote. “As reviewed here, emerging evidence supports the potential of texture analysis as a neuroimaging marker of Alzheimer’s disease.”
There have been a slew of studies investigating MR texture analysis’ ability to classify and predict Alzheimer’s, as well as positive findings related to differentiating it from other dementias.
One such study found that combining the approach with cerebrospinal fluid data could improve accuracy in classifying the disease. Another analysis calculated texture features from datasets and confirmed dementia with Lewy bodies in 70% of patients suffering from cognitive decline. It also separately diagnosed 91.7% of dementia patients with Alzheimer’s disease, researchers reported.
Additionally, the authors pointed to a study that found MRTA hippocampal texture features more accurately predicted a transformation from mild-cognitive impairment to Alzheimer’s within two years, compared to using hippocampal volume. This supports the idea that texture captures more information than commonly measured volumes.
Along with this potential comes plenty of obstacles to widespread clinical adoption of texture analysis, the researchers noted. And many are related to a lack of standardization.
For example, acquisition protocols are very important in generating texture features, but are highly variable among T1-weighted images.
Another key barrier is related to image intensity discretization, which involves resampling image intensity values. There are two options for achieving this—using a set number of discrete values or fixed intensity resolutions—but researchers don’t yet know which is more important.
The lack of standardized feature-extracting methods and modeling also makes it difficult to compare results across various studies, the authors wrote.
More than 47 million people currently live with dementia, and that figure is expected to triple by 2050, the authors wrote. The global cost of care for these patients is predicted to top $800 billion, according to a 2015 estimate, and that figure is only likely to grow further.
The “new era” that Jia-Hui-Cai et al. described will require more research and prospective multi-center studies to usher in the clinical adoption of texture analysis. Future approaches should seek to incorporate radiomic features and genomic data and employ deep learning trained on texture data to classify and predict Alzheimer’s disease, the group concluded.