Virtusa / vLife / AI as a Service

PREDICTING TYPE 2 DIABETES

Type 2 Diabetes is a chronic disease and is a major health concern across the globe. In this proof of concept (POC), we are using medications and labs data to predict whether a patient has Type 2 Diabetes.

PREDICTING TYPE 2 DIABETES

Type 2 Diabetes is a chronic disease that is a major health concern across the globe. The number of people with diabetes has risen drastically in the last few decades. The disease is not only a huge financial burden but also a major cause of blindness, kidney failure, heart attacks, stroke, and lower limb amputation. The WHO projects that diabetes will be the seventh leading cause of death in 2030.

In this POC, we are attempting to predict T2D from the laboratory results captured for a patient. We are using a National Health and Nutrition Examination Survey (NHANES) to predict Type 2 Diabetes from laboratory test results data.

In the POC, we are using medications and labs data. The medications data contains diagnostics details where we select the patients with T2D. The labs data includes laboratory measures that will be our predictors. We are using AdaBoost to train the model and rank the most predictive features.

Try Demo

 

AUTOMATIC BRAIN TUMOR SEGMENTATION (ABTS)

Brain tumor segmentation seeks to separate healthy tissue from tumorous regions. The most common diagnostic tool used to identify brain tumors is the MRI.

AUTOMATIC BRAIN TUMOR SEGMENTATION (ABTS)

Brain tumor segmentation seeks to separate healthy tissue from tumorous regions. The most common diagnostic tool used to identify brain tumors is the MRI. It is noninvasive and can image diverse tissue types and physiological processes. A single patient will produce more than 600 images from a single MRI. Because manual segmentation is slow and tedious, there is a high demand for computer algorithms that can do it quickly and accurately, allowing for more immediate patient care and higher throughput treatment times. ​​

ABTS has the potential to decrease lag time between diagnostic tests and treatment by providing an immediate report of tumor location. The algorithm uses convolutional neural networks (CNNs) for semantic segmentation of images. An ABTS algorithm can distinguish between tumor and healthy tissue, actively enhancing tumor and non-advancing tumor regions. The enhanced images let clinicians focus more on patient care and improve their chances of survival. ​

Download Brochure | Request Demo

PREDICTING FREQUENT OPIATE PRESCRIBERS

Over prescription of opioids is an epidemic that has caused significant amount of deaths in the United States. In this POC, we are using prescriber info to predict frequent opiate prescribers.

PREDICTING FREQUENT OPIATE PRESCRIBERS

Fatal drug overdoses account for a significant portion of accidental deaths in adolescents and young adults in the United States. Most fatal drug overdoses involve opioids. Medically, they are used as powerful pain relievers; however, they also produce feelings of euphoria. This makes them highly addictive and prone to abuse.

Over the past 15 years, deaths from prescription opiates have quadrupled, but so has the number of opiates prescribed.

In this POC, we are predicting the likelihood that a given doctor is a significant prescriber of opiates. We are using CMS.gov’s prescriber information data present in vLife™ and utilizing NPI, gender, prescriber’s credentials, and specialty as predictors.

Try Demo

Predicting Parkinson’s Disease

Parkinson’s disease (PD) is a progressive nervous system disorder that affects movement. One of the early symptoms is speech changes. In this proof of concept (POC), we are using patient voice features to predict Parkinson’s disease.

Predicting Parkinson’s Disease

PD is a progressive nervous system disorder that affects movement. Symptoms start gradually, sometimes with a barely noticeable tremor in just one hand or speech changes. Although PD can’t be cured, medications might significantly improve your symptoms. PD signs and symptoms can be different for everyone. Early signs may be mild enough to go unnoticed and may include tremor, slowed movement, rigid muscles, impaired posture and balance, and speech and writing changes. PD is the second most common neurodegenerative disorder in the United States following Alzheimer’s disease. The burden of chronic conditions such as PD is projected to grow substantially over the next few decades as the size of the elderly population grows.

In this proof of concept, we are using the Logistic Regression classifier to diagnose whether a patient is suffering from PD using his or her voice features. Research suggests that speech production can be modeled as a nonlinear dynamical system, wherein small perturbations in the interaction of its parts give rise to chaotic yet deterministic behavior.

We are using the Parkinson Multiple Sound Recording dataset available in vLife™ to predict PD. The data belongs to 20 patients with PD (6 males and 14 females) and 20 healthy subjects (10 male and 10 female). Multiple types of sound recordings are taken from each subject. Praat software is used to calculate features like shimmer, jitter, harmonic to noise (HTN), and pitch from each subject’s voice.

Have a question about vLife?

Contact us

    Yes, I want Virtusa to keep me up-to-date with recent industry developments including insights, upcoming events, and innovative solution capabilities according to the privacy policy
  • This field is for validation purposes and should be left unchanged.