ANDREW LIANG

INVESTOR PITCH

​

What is your innovative product/service?

​

Communication is an important asset to human function that dictates the bare elements of life. With 7.5 million deaf individuals, 50 million individuals over 65, and 4 million newborns born each year, auditory impediments complicate fundamental communication. Furthermore, this verbalization is a primary aspect of professional medical diagnosis. Without the appropriate diagnosis, symptoms can go untreated, festering into long lasting conditions. This can ultimately be detrimental to not only the patient’s health but also their pockets. As such, we designed Specula, an automated approach for diagnosis of neurodegenerative diseases in infants, the elderly, and the deaf. Utilizing nonverbal behaviors from the autonomic nervous system like sporadic eye movements (which has linked to diseases like Alzheimer’s, Parkinson’s, and autism), our approach harnesses industry standard machine learning (ML) models to analyze and assess proprietarily collected data using eye tracking virtual  reality glasses and correlate it with the aforementioned diseases. Our VR glasses demo a typical eye test (a rapidly moving black dot on a white background) while simultaneously recording eye movements through a camera appendage. Vectorizing the eye movements from open-source APIs provides the user with a custom made data sample. Comparing the user’s results with other data samples through the ML model allows for a computational analysis predicting the likelihood of certain neurodegenerative diseases. Since Specula’s backbone is foundationally code, the technology can be extrapolated into the context of an iOS or Android app on a cellular device. Leveraging the camera’s capabilities to capture live feed eye data, users can opt for quick diagnosis within their fingertips. Given that our product’s utility is corroborated with leading scientific research from Nature and SciHub, Specula can ramp up medical diagnosis by utilizing artificial intelligence based on models - at home and in the hospital - and offer an alternative approach to cross validate with human diagnosis, subject to natural variability. 

​

What challenge(s) is your product/service designed to solve?

​

Every year approximately over 50 million Americans are affected by more than 600 different neurodegenerative diseases. With the population of the United States rising and aging, there is an increasing number of Americans affected by neurodegenerative diseases, most commonly Alzheimer’s, Parkinson’s, and multiple sclerosis. These diseases are the product of progressive neural degeneration leading to central or peripheral nervous system dysfunction. That is reflected in changes in memory, cognitive capabilities, motor control, and other symptoms. Financially, these diseases cost the U.S. economy billions of dollars each year through health care costs and opportunity costs. Meek et al. estimated in 1998 that Alzheimer’s alone costs America $100 billion per year. This does not account for the intense emotional burden on patients and their families.

 

Presently, neuron regeneration is impossible so modern therapies largely target patients with moderate to severe symptoms with an intent to slow disease progression. However, much of the time at this stage, the neurodegeneration has reached the point of irreversible damage. Therefore, early intervention mechanisms are key. Noble and Burns estimate that interventions to delay disease progression by a year would reduce the number of predicted cases of Alzheimer’s by 9.2 million. This provides two significant advancements. First, it would allow for early diagnosis and treatment. Second, this enables scientists to track disease progression which can be used to assess the efficacy of different treatments which will be key in future drug development (Nobel and Burns, 2010). 

 

Unfortunately, precise determinations and differentiation of neurodegenerative diseases is difficult using traditional biomarkers given the complexity and similarities of protein interactions amongst different neurodegenerative diseases. This is further complicated by neuroinflammation and a tendency for diseases to co-occur. Indeed, approximately 80% of cases require confirmation postmortem through an examination of brain tissue (Mok et al., 2004).

 

What are the key features of your product/service that make it special?

​

Specula’s core features include vectorized eye tracking using deep computer vision technologies. By locating and comprehending a given patient’s sporadic eye movements, Specula will be able to diagnose and generate a correlation of correlated diseases such as Alzheimer’s, Parkinson’s, autism, etc. Specula initially requests the patient to conduct a reading test. During the reading test, Specula utilizes computer vision tools such as open-source libraries to locate eye movement by detecting standard change in pixel channel values and cache the resulting bounding coordinates around the eye per frame to a backend system. Using deep-based convolutional neural networks (CNNs), Specula analyzes the stored coordinates from the test and draws a prediction for a potential neurodegenerative disease. Through such a method, cost-time efficiency, accessibility, and accuracy is optimally considered. With proprietary data collected from sources such as virtual reality headsets, Specula’s deep learning model will be able to constantly train and deploy a working model to a cloud server for patients to easily access and test robustly.

 

Specula is also capable of analyzing other neurological data relating to memory that correlates to brain disorders and impacts. By having interactive quests, Specula will be able to gather patient results, further conduct diagnoses, and gamify the onboarding process. Using computational algorithms, Specula can draw correlations and be capable of directly assessing probabilities and predictions of future neurodegenerative diseases for a given patient. This novel approach and application to involve patients with Specula’s services will lead to an increase in the interaction between the patient and the app, which will allow for greater amounts of patient data, further providing for greater accuracy in our machine learning model.
 

How is your product/service innovative and different from other products/services intended to solve the same challenges?

 

Specula is a one of a kind product, incorporating data collection and analysis into one unit. From a medical perspective, the use of ML to neurodegenerative diseases has a literature base limited to the autism diagnosis. Thus, our project is unique in its application of the same principles of sporadic eye movement and ML to different neurodegenerative diseases. Designed off traditional VR glasses, Specula’s carefully positioned camera and mirror system allows for eye movement data to be tracked with proper lighting and without interfering with the patient’s visual range. As such, Specula itself is a utility for data collection. Once large sums of data are calculated through Specula, we can use that proprietary data for our custom-built machine learning models. The model architecture is based on open-sourced RESNET techniques, but the model output will be proprietary to our specific data. The trained model is then installed into the firmware of the Specula glasses and can be used for diagnosis. Because our product is electronically based, we can incorporate software updates allowing the same glasses to be compatible for new diseases once those new models are trained. In addition, we can send updates with higher fidelity machine learning models. This entire automation process would be unique to Specula. Current diagnosis approaches necessitate human interaction with the data, either through recorded data or hands-on procedures. However, with Specula, the procedure is done through software, eliminating the need for expensive diagnosis by doctors. Additionally, we believe our software can outperform human diagnosis. Machine Learning models are experts at detecting subtle variations in data and honing in on these differences to make quantitative predictions that are more concrete than human measurements. Ultimately, we believe Specula will revolutionize the way medical diagnoses are performed and will unfold undiscovered relationships between diseases and eye movements.

​