Research News

Cleveland Clinic estimates up to 50% of adults aged 85 and older are at risk of developing dementia. Although dementia cannot be cured, some forms can be treated and even reversed by addressing a root cause. Cleveland Clinic is working to identify individuals with one of these causes, normal pressure hydrocephalus (NPH), through deep learning and gait analysis. 

What is normal pressure hydrocephalus?  

NPH is a condition with an abnormal buildup of cerebrospinal fluid in the brain. The buildup manifests with enlargement of the fluid-filled spaces in the brain, called ventricles, and changes brain function leading to gait impairment, urinary incontinence, memory loss and other cognitive issues. NPH affects approximately 8.4 million people over age 80 worldwide and is the cause of up to 5% of dementia cases.  

“The most common symptom of NPH is gait impairment, but gait is also affected by structural issues like spinal stenosis and arthritis, or other neurogenerative conditions like Parkinson’s or Alzheimer’s,” says James Liao, MD, Associate Staff, Cleveland Clinic Center for Neuro-Restoration. “This leads to NPH being undiagnosed or misdiagnosed.” 

Treatment and diagnosis  

Diagnosing NPH depends on identifying the fluid buildup using brain imaging and performing a diagnostic procedure called a “drain trial” to remove some fluid to see if symptoms like gait impairment improve. Only patients whose symptoms improve are recommended to have surgery for a permanent shunt that continuously drains fluid from the brain to the abdomen.  

Dr. Liao says current diagnostic methods are not always accurate in detecting which patients have NPH and would benefit from surgery.  

“In one previous study, 22 of 38 patients who underwent the drain trial did not see symptoms improve,” he explains. “However, as part of the study, all patients received the shunt surgery anyway. It was revealed that over half of those patients who technically failed the diagnostic test would still benefit from surgery.”  

Even though this was a small study, Dr. Liao says it demonstrates the need for a more accurate method of diagnosis.  

Gait impairment as a diagnostic tool 

The most common side effect of NPH is gait impairment, the deterioration of an individual’s ability to walk. Clinicians currently take videos of patients as they walk over a set distance to monitor gait performance, then they manually calculate a few metrics like step count or speed. For years, clinicians have been interested in using more advanced gait metrics to diagnose NPH, or to predict which patients would respond to the shunt surgery, Dr. Liao says.  

Traditionally, these advanced gait metrics would only come from complex motion capture systems that involve wearing special optical markers, which are not feasible to use in real-world clinical medicine. So, these metrics were out of reach, until recently. Now, computer vision techniques have improved enough to extract complex gait metrics from gait videos alone. 

The research team has begun to collect videos of NPH patients walking before and after their diagnostic drain trials, using a special hallway outfitted with several cameras mounted at different angles. Once the videos are collected, the team will use deep learning algorithms to de-identify the videos and extract movement information. These algorithms, developed by IBM Research, will convert the gait videos into clinically relevant gait metrics, as well as feature representations for a broad range of gait-analysis applications. 

The extracted movement information will then be used to train machine learning algorithms on tasks like, detecting the minute differences in gait that distinguish patients with and without NPH and predicting patients who would respond to the shunt surgery. 

“These computational techniques have the potential to not only help advance our knowledge of NPH but to support future research into other neurological disorders such as cerebral palsy, Parkinson’s, Alzheimer's, or other dementia,” Dr. Liao says. “In fact, the new Neurological Institute building will have a dedicated gait video capture area that patients will pass through as they check-in.”  

Dr. Liao continues, “This goes beyond neurology too. Gait changes predict fall risk, functional ability, mortality and hospitalization risk. If we can measure and analyze gait seamlessly, we may eventually consider gait metrics to be vital signs just like heart rate or blood pressure.”