Matt Hoffman

This article was originally published on Diagnostic Imaging's sister site 

Pairing MRI with a new artificial intelligence (AI) based deep learning platform called DystoniaNet can identify patients with dystonia in under a second, new research has revealed.

According to investigators from Mass Eye and Ear in Boston, this tool can help providers truncate the time to diagnosis for this condition which causes involuntary muscle contractions that prompt abnormal movements and postures. Correctly identifying dystonia can sometimes take up to 10 years. They published their findings on Sept. 28 in 

Proceedings of the National Academy of Sciences (PNAS).

“We basically leveraged the advances made in deep learning and designed an architecture that was able to look at raw structural MRI and find a biomarker for dystonia that could help with the diagnosis of this disorder,” study author Davide Valeriani, PhD, postdoctoral research fellow in the Dystonia and Speech Motor Control Laboratory at Mass Eye and Ear and Harvard Medical School told 

. “We specifically took an approach that could be easily clinically translated, which is why we used raw structural MRI. This kind of measure is what comes out directly from the scanner and is available to many clinicians in-clinic.”

All told, when doing a comparison using MRI data from 612 individuals—220 healthy subjects and 392 with three types of isolated focal dystonia (laryngeal dystonia, n = 279; cervical dystonia, n = 59; blepharospasm, n = 54)—the DystoniaNet platform diagnosed the condition with 98.8-percent accuracy in a matter of 0.36 seconds. Specifically, it showed 98.2-percent accuracy in diagnosing laryngeal dystonia, 100 percent in diagnosing cervical dystonia, and 98.1 percent in diagnosing blepharospasm, while referring 3.5 percent of patients (n = 6) for further examination.

Given the platform's performance and the average length of time to diagnosis, it has potential to beneficial impact the movement disorder field, the team said. 

Valeriani explained that the DystoniaNet platform takes the MRI and learns from the training algorithm developed from one of the largest dystonia datasets, and then uses this data-driven approach to identify the condition. It was developed over the last several years by Valeriani and co-author Kristina Simonyan, M.D., Ph.D., DrMed, director of laryngology research at Mass Eye and Ear, and was previously assessed in an analysis-driven marker evaluation.

“[The analysis-driven] results were quite encouraging, where we would have an accuracy between 71 percent and 81 percent correctly diagnosing dystonia, and research from other labs also has shown that similar analysis-driven biomarkers can be implemented for diagnosing dystonia” said Simonyan, who is also an associate neuroscientist at Massachusetts General Hospital and associate professor of otolaryngology, head and neck surgery at Harvard Medical School. “However, that was not good enough for us because that type of paradigm has to involve substantial data processing, which has been clinically challenging to do. Us clinicians just don’t have the luxury of several hours per patient to process their data. In a way, our study was designed to lessen the time burden on the clinician and to speed up the diagnosis.”

The performance of the biomarker and its DystoniaNet pipeline was consistently stable independent of the magnetic field strength (accuracy range, 98–100), MRI scanner vendor (accuracy range, 96.9–100), head coil (accuracy range, 95.2–100), T1-weighted image acquisition sequence (accuracy range, 98.3–100), or a data collection site (accuracy range, 97.6–100). The specificity was additionally confirmed with a third and supplementary independent dataset of 1480 healthy controls (accuracy, 96.9 percent; referral rate, 2.6 percent).

Simonyan noted that traditionally, this diagnosis is made based on clinical observations, and available literature suggests that that the agreement on dystonia diagnosis between clinicians using purely clinical assessments is as low as 34 percent, and with roughly 50 percent of the cases going misdiagnosed or under-diagnosed at a first visit.

It is estimated that 35 of 100,000 individuals have isolated or primary dystonia, though it is believed that this prevalence is underestimated due to the current challenges in diagnosis. Although Parkinson's disease or stroke can lead to the development of dystonia, the majority of isolated dystonia cases have no known cause and impact only a single muscle group in the body. Focal dystonia can lead to disability and problems with physical and emotional quality of life.

“In some sense, this provided us an opportunity to train this machine-learning algorithm to diagnose dystonia because we had a large available dataset of neuroimages,” Valeriani said. “But, we also made the development of DystoniaNet open to the possibility of extending it to other disorders. For example, we took away all the prognosis that’s required and just stared with raw structural MRI. Every patient could get that, and we have this data for many different disorders."

Currently, he said, the team is testing this on some other disorders with very promising results with approximately 80-percent accuracy. The goal is to determine how much the architecture will generalize to other disorders. 

For more coverage based on industry expert insights and research, subscribe to the Diagnostic Imaging e-newsletter

Articles

The platform, dubbed DystoniaNet, was able to identify 3 varieties of focal dystonia in a matter of 0.36 seconds with almost 100-percent accuracy.

AI Identifies Focal Dystonia Almost Instantaneously on MRI

This article was originally published on Diagnostic Imaging's sister site, NeurologyLive.

The U.S. Food & Drug Administration (FDA) has approved the first drug designed to help image a distinctive characteristic of Alzheimer’s disease – the radiopharmaceutical called flortaucipir F18.

Marketed as Tauvid and developed by Avid Radiopharmaceuticals, this radioactive diagnostic agent is indicated for use with PET imaging of the brain. It helps providers estimate the density and distribution of aggregated tau neurofibrillary tangles (NFTs), a primary marker of Alzheimer’s disease.The safety and effectiveness of the imaging agent were evaluated in a pair of clinical studies, in which each had 5 evaluators read and interpret the imaging.

“Alzheimer disease is a devastating condition that affects millions of Americans. This approval will provide health care professionals with a new type of brain scan to use in patients being evaluated for Alzheimer disease,” said Charles Ganley, MD, director, Office of Specialty Medicine, Center for Drug Evaluation and Research, FDA, in a statement. “While there are FDA approved imaging drugs for amyloid pathology, this is the first drug approved for imaging tau pathology, 1 of the 2 neuropathological hallmarks of Alzheimer disease, and represents a major advance for patients with cognitive impairment being evaluated for the condition.”

The FDA noted that as of now, Alzheimer disease can only be definitively diagnosed by postmortem pathologic evaluation of a patient’s brain, and 3 imaging agents are approved for postmortem amyloid pathology with PET scans.
	
	

NIH Data Suggest Brain Glucose Metabolism Is Linked to Alzheimer Pathogenesis

The data from the 2 clinical trials assessing the flortaucipir F18 tracer included more than 300 patients, including those with terminal illness who agreed to participate in a postmortem brain donation program, as well as those with mild cognitive impairment (MCI). Those findings suggested that the tau tracer can support a pathological diagnosis of Alzheimer disease by identifying the underlying presence of NFTs at the B3 level (Braak stages V to VI) and high levels of Alzheimer disease neuropathologic change (ADNC) per National Institute on Aging–Alzheimer Association (NIA–AA) criteria.2,3

The first study, Study A16, enrolled 156 terminally ill patients, of which 64 died within 9 months of their brain scan. The reading of those scans was compared to postmortem readings from independent pathologists who evaluated the density and distribution of NFTs in the same brain. The data suggested that the flortaucipir F18 images had a high probability of correctly evaluating patients with tau pathology and had an average-to-high probability of correctly evaluating patients without tau pathology.2

The second study, the FR01 Validation Study, included those from the first study, in addition to 18 additional patients with terminal illness and 159 patients with cognitive impairment who were being assessed for Alzheimer disease. This assessment judged agreement between evaluations from each clinician scan reader, with perfect reader agreement rated as 1 and no reader agreement as 0. Data showed that reader agreement was 0.87 across all 241 patients.1

A separate subgroup analysis with the 82 terminally ill individuals with or without dementia diagnosed postmortem and the 159 patients with cognitive impairment showed that reader agreement was 0.90 for the patients in the indicated population and 0.82 in the terminally ill patients.1 All told, B3 level NFTs sensitivity was 89.1% (95% CI, 77.0–95.3) and specificity was 86.1% (95% CI, 71.3–93.9), and high Alzheimer disease neuropathologic change (ADNC) level sensitivity was 95.1% (95% CI, 83.9–98.7) and specificity was 82.9% (95% CI, 68.7–91.5).3

Of the 64 patients in the primary cohort from A16, the mean age was 82.5 (standard deviation [SD], 9.6) years. In total, 77% (n = 49) had dementia, 2% (n = 1) had MCI, and 22% (n = 14) had normal cognition. Among the 156 enrolled participants, 9% (n = 14) experienced at least 1 treatment-emergent adverse event (AE).

The scans in the subgroup analysis predicted a B3 level of tau pathology with sensitivity ranging from 92.3% (95% CI, 79.7–97.3) to 100.0% (95% CI, 91.0–100.0) and specificity ranging from 52.0% (95% CI, 33.5–70.0) to 92.0% (95% CI, 75.0–97.8). Elevated levels of ADNC was predicted with sensitivity of 94.7% (95% CI, 82.7–98.5) to 100.0% (95% CI, 90.8–100.0) and specificity of 50.0% (95% CI, 32.1–67.9) to 92.3% (95% CI, 75.9–97.9).
	
	For that full cohort data set, authors Adam S. Fleisher, MD, MAS, et al. noted that all 5 readers met the success criteria, defined as having ≥3 of the 5 readers above the lower bounds of the 95% CI for both sensitivity and specificity of 50% or greater. For NFTs, sensitivity ranged from 89.1% (95% CI, 77.0–95.3) to 93.5% (95% CI, 82.5–97.8) and specificity from 66.7% (95% CI, 50.3–79.8) to 94.4% (95% CI, 81.9–98.5). For a high level of ADNC, sensitivity ranged from 95.1% (95% CI, 83.9–98.7) to 97.6% (95% CI 87.4–99.6) and specificity from 65.9% (95% CI, 50.5–78.4) to 90.2% (95% CI, 77.5–96.1).

The FDA noted that the most common AEs in patients using the tracer were headache, injection site pain, and increased blood pressure. It is not indicated for use in the evaluation of patients for chronic traumatic encephalopathy.1

REFERENCES
	1. FDA Approves First Drug to Image Tau Pathology in Patients Being Evaluated for Alzheimer’s Disease. News release. FDA; May 28, 2020. Accessed May 28, 2020. https://www.fda.gov/news-events/press-announcements/fda-approves-first-drug-image-tau-pathology-patients-being-evaluated-alzheimers-disease
	2. Clark CMï»¿, Pontecorvo MJï»¿, Beach TGï»¿, et al; AV-45-A16 Study Group. Cerebral PET with florbetapir compared with neuropathology at autopsy for detection of neuritic amyloid-Î² plaques: a prospective cohort study. Lancet Neurol. 2012;11(8):669-678. doi:10.1016/S1474-4422(12)70142-4
	3. Fleisher AS, Pontecorvo MJ, Devous Sr MD, et al. Positron Emission Tomography Imaging With [18F]flortaucipir and Postmortem Assessment of Alzheimer Disease Neuropathologic Changes. JAMA Neurol. Published online April 27, 2020. doi:10.1001/jamaneurol.2020.0528

For use with PET imaging, the radiopharmaceutical helps estimate density and distribution of tau, a marker of Alzheimer's disease.

Matt Hoffman

Articles

AI Identifies Focal Dystonia Almost Instantaneously on MRI

FDA Approves First Radiopharmaceutical for Imaging Tau