DystoniaNet Diagnosis for Dystonia

Overseen By Kristina Simonyan, MD, PhD

Age: Any Age

Sex: Any

Trial Phase: Academic

Sponsor: Massachusetts Eye and Ear Infirmary

Disqualifiers: Inability to consent, MRI restrictions, pregnancy, others

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Compensation: Varies

Number of Visits: Unknown

1000 Participants Needed

What You Need to Know Before You Apply

What is the purpose of this trial?

This trial tests a new tool called DystoniaNet, which uses deep learning (a type of artificial intelligence) to diagnose dystonia, a condition that causes muscle contractions and movement problems. The goal is to determine if DystoniaNet can quickly and accurately identify dystonia in patients. The trial will compare results from individuals with dystonia, other movement disorders, and some non-neurological conditions that show similar symptoms. Individuals who have experienced symptoms of dystonia or related movement issues, such as frequent muscle spasms, might be a good fit for this study. As an unphased trial, this study offers a unique opportunity to contribute to groundbreaking research in diagnosing movement disorders.

Do I need to stop my current medications for the trial?

The trial information does not specify whether you need to stop taking your current medications.

What prior data suggests that the DystoniaNet deep learning platform is safe for diagnosing dystonia?

Research has shown that DystoniaNet, a smart computer program, is being developed to help diagnose dystonia, a condition where muscles contract involuntarily. This program uses artificial intelligence to analyze MRI scans and accurately detect signs of dystonia. One study demonstrated that DystoniaNet identified dystonia cases with 98.8% accuracy from MRI scans.

Regarding safety, DystoniaNet is neither a drug nor a device used directly on patients. It functions as a smart tool to assist doctors in making diagnoses. Consequently, there are no safety concerns or side effects associated with using DystoniaNet. It simply analyzes existing medical images to aid in diagnosing dystonia. ^{1 2 3 4 5}

Why are researchers excited about this trial?

Researchers are excited about the DystoniaNet trial because it uses advanced technology to diagnose dystonia accurately and quickly. Unlike traditional methods that rely heavily on subjective clinical evaluations and can be time-consuming, DystoniaNet aims to provide an objective and fast diagnosis using innovative machine learning techniques. By improving the speed and accuracy of dystonia diagnosis, DystoniaNet has the potential to significantly enhance patient care and treatment outcomes.

What evidence suggests that the DystoniaNet platform is effective for diagnosing dystonia?

Research has shown that DystoniaNet, an advanced computer program, excels at diagnosing dystonia, a movement disorder. One study found that DystoniaNet identified dystonia from MRI scans with 98.8% accuracy, nearly 20% better than other similar methods. Typically, only about 5% of patients receive a correct dystonia diagnosis initially, with many waiting over 10 years for an accurate diagnosis. This trial will include a prospective clinical validation arm to test DystoniaNet's performance in real-world settings and a retrospective validation arm to compare its diagnostic performance against normal neurological states and other conditions. DystoniaNet could significantly improve the speed and accuracy of dystonia diagnosis, potentially transforming how dystonia is identified and treated. ^{2 3 4 5 6}

Who Is on the Research Team?

Kristina Simonyan, MD, PhD

Principal Investigator

Massachusetts Eye and Ear

Are You a Good Fit for This Trial?

This trial is for individuals with various forms of dystonia or conditions that resemble dystonic symptoms, such as Parkinson's disease and essential tremor. It includes people of all ages, genders, and ethnic backgrounds. Those who can't give consent or have MRI-incompatible body modifications or devices are excluded.

Inclusion Criteria

I have a type of dystonia, such as in my neck, eyes, jaw, hand, or it affects my whole body.

I have a type of dystonia, such as in my neck, eyes, jaw, hand, or it's more widespread.

I have a condition like Parkinson's, essential tremor, or another similar disorder.

Exclusion Criteria

You cannot have an MRI of your brain because you have certain tattoos or metal objects inside your body that cannot be removed, or because you are pregnant or breastfeeding.

I am unable to understand and give consent for my treatment.

Timeline for a Trial Participant

Screening

Participants are screened for eligibility to participate in the trial

2-4 weeks

Retrospective Study

Retrospective studies will clinically validate the diagnostic performance of DystoniaNet compared to a normal neurological state and other conditions.

4 years

Prospective Study

Prospective randomized studies will validate DystoniaNet performance for accurate, objective, and fast diagnosis of dystonia in the clinical setting.

4 years

Follow-up

Participants are monitored for safety and effectiveness after diagnosis using the DystoniaNet algorithm.

4 weeks

What Are the Treatments Tested in This Trial?

Interventions

• DystoniaNet

Trial Overview

The study is testing DystoniaNet, a deep learning platform designed to diagnose isolated dystonia by analyzing medical data. The research will look back at past cases and also include new patients to validate the accuracy of this diagnostic tool.

How Is the Trial Designed?

Group I: Prospective clinical validation of DystoniaNet Experimental Treatment 1 Intervention

Prospective randomized studies will validate DystoniaNet performance for accurate, objective, and fast diagnosis of dystonia in the actual clinical setting.

Group II: Retrospective clinical validation of DystoniaNet Active Control 1 Intervention

Retrospective studies will (1) clinically validate the diagnostic performance of DystoniaNet compared to a normal neurological state (normative test), and (2) develop and test DystoniaNet extensions in comparison with other neurological and non-neurological conditions (differential test).

Who Is Running the Clinical Trial?

Massachusetts Eye and Ear Infirmary

Lead Sponsor

Trials

115

Recruited

15,000+

Published Research Related to This Trial

A new neural network-based software has been developed to objectively measure and recognize symptoms of blepharospasm (BSP), including blinks and spasms, which can help standardize the assessment of this complex condition.

The software demonstrated high sensitivity for detecting brief and prolonged spasms, making it a promising tool for clinicians to accurately evaluate BSP severity compared to traditional methods.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A neural network-based software to recognise blepharospasm symptoms and to measure eye closure time. Trotta, GF., Pellicciari, R., Boccaccio, A., et al.[2020]

This study demonstrates that using smartphone-coupled inertial sensors and machine learning can provide objective, home-based assessments of dystonia severity in children with dyskinetic cerebral palsy, achieving average F1 scores of 0.67 for lower and 0.68 for upper extremities.

The results suggest that these automated assessments could complement traditional clinical evaluations, offering more frequent and reliable monitoring of dystonia, although further research is needed to improve model accuracy and data collection.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Home-Based Measurements of Dystonia in Cerebral Palsy Using Smartphone-Coupled Inertial Sensor Technology and Machine Learning: A Proof-of-Concept Study. den Hartog, D., van der Krogt, MM., van der Burg, S., et al.[2022]

A deep learning model was able to accurately classify individual features of hyperkinetic seizures, achieving an F1 score of 0.84 for detecting emotional signs and 0.83 for identifying dystonia, based on a dataset of 38 seizure videos from 19 patients.

This study demonstrates the potential of using advanced deep learning techniques to automate the analysis of seizure characteristics, which could enhance the understanding and treatment of epilepsy.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Automated video analysis of emotion and dystonia in epileptic seizures. Hou, JC., Thonnat, M., Bartolomei, F., et al.[2022]

A deep learning algorithm called DystoniaNet was developed to diagnose isolated dystonia from brain MRIs, achieving an impressive accuracy of 98.8% across 612 subjects, including 392 patients with different forms of dystonia and 220 healthy controls.

DystoniaNet identified specific brain regions as biomarkers for dystonia, significantly improving diagnostic agreement among clinicians from 34% to a much higher level, thus enhancing clinical decision-making and potentially reducing misdiagnosis and delays in treatment.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A microstructural neural network biomarker for dystonia diagnosis identified by a DystoniaNet deep learning platform. Valeriani, D., Simonyan, K.[2021]

Dystonia is a complex movement disorder characterized by involuntary, repetitive movements and abnormal postures, which can be challenging to diagnose due to its similarity across various causes.

The review proposes an eight-question diagnostic approach to help clinicians, both specialists and general neurologists, effectively identify dystonic syndromes and determine appropriate investigations for accurate diagnosis.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Diagnosis of dystonic syndromes--a new eight-question approach. Bertram, KL., Williams, DR.[2022]

A study of 1,618 participants with isolated non-focal dystonia used data-driven clustering to identify common patterns of dystonia affecting various body regions, revealing strong associations between bilateral upper and lower limbs, which may indicate a new subtype of dystonia.

The analysis highlighted three major clusters of dystonia: cervical dystonia with nearby regions, bilateral hand dystonia, and cranial dystonia, reinforcing the idea of segmental patterns in dystonia affecting the cranial and cervical areas.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Anatomical categorization of isolated non-focal dystonia: novel and existing patterns using a data-driven approach. Younce, JR., Cascella, RH., Berman, BD., et al.[2023]

Citations

1.

simonyanlab.meei.harvard.edu

simonyanlab.meei.harvard.edu/research/dystonianet-a-deep-learning-platform-for-dystonia-diagnosis-and-treatment/

DystoniaNet: A Deep Learning Platform for Dystonia ...

We have developed the first objective, accurate, fast and cost-efficient deep-learning platform, DystoniaNet, for dystonia diagnosis.

2.

clinicaltrials.gov

clinicaltrials.gov/study/NCT05317390

Clinical Validation of DystoniaNet Deep Learning Platform ...

It is estimated that only 5% of patients receive an accurate diagnosis at symptom onset, and the average diagnostic delay extends up to 10.1 years. This study ...

3.

pubmed.ncbi.nlm.nih.gov

pubmed.ncbi.nlm.nih.gov/33004625/

A microstructural neural network biomarker for dystonia ...

DystoniaNet significantly outperformed shallow machine-learning algorithms in benchmark comparisons, showing nearly a 20% increase in its ...

4.

masseyeandear.org

masseyeandear.org/news/press-releases/2020/09/new-artificial-intelligence-platform-uses-deep-learning-to-diagnose-dystonia

New Artificial Intelligence Platform Uses Deep Learning to ...

New study showed DystoniaNet AI platform detected cases of dystonia from an MRI with 98.8 percent accuracy. There is currently no diagnostic ...

5.

amazon.science

amazon.science/latest-news/amazon-research-award-recipient-develops-new-tool-to-diagnose-dystonia

Amazon Research Award recipient develops new tool to ...

Amazon Research Award recipient develops new tool to diagnose dystonia. Dr. Kristina Simonyan and her team created an AI-based deep learning platform that ...

6.

pnas.org

pnas.org/doi/10.1073/pnas.2009165117

A microstructural neural network biomarker for dystonia ...

We developed a deep learning algorithmic platform, DystoniaNet, to automatically identify and validate a microstructural neural network ...