Deep Learning Technology

Marpai is a technology company bringing the deep learning revolution into the self-funded health plan sector to save lives, improve lives and radically reduce healthcare spending.

Deep Learning. The Most Advanced Form of AI.

Marpai takes a preventive healthcare approach to improving outcomes and reducing healthcare costs using the world’s first and only purpose-built, deep learning health prediction framework.

Deep Learning, the most advanced AI, delivers the greatest form of health state prediction available today—allowing Marpai to predict near-term health events to prevent chronic illness (like Type 2 Diabetes) and major procedures (like knee surgery)  before they happen which protects health plan members well beyond traditional preventive health solutions.

Powered by a deep neural network brain that mimics the logic and learning of the human brain, Marpai’s deep learning platform anticipates and prevents health events with unmatched speed and accuracy. We stop chronic illness before it develops, identifying health risks with maximum accuracy.

To stay ahead of the developing health issues and prevent unknown risks from developing, the predictive power of Marpai’s deep learning-based solution is a necessity.

What is Deep Learning

Deep learning is the most advanced subset of AI, leveraging deep neural networks that take inspiration from how the human brain works.

Deep Learning is an artificial intelligence method that imitates the way the human brain works in the sense of processing data and creating patterns for use in decision making. It utilizes a hierarchical level of artificial neural networks to carry out the learning process involved in machine learning. The artificial neural networks are built like the human brain, with neuron nodes connected like an interconnected web.

As more data is fed into the neural network, it becomes better at intuitively understanding the meaning of new data. This allows it to predict and prevent increasingly advanced threats such as health risks. It does not require a (human) expert to help it understand the significance of new features.

The advantage of deep learning over other forms of machine learning is the end-to-end processing of data. These three factors contribute to deep learning’s greater level of accuracy.

  • Elimination of the feature engineering phase, an inherent part of machine learning, which involves a human expert identifying and selecting the features for analysis.

  • Analysis of all the available data in the training sample.

  • Encompassing of representation learning the ability of a model to get input low level features (such as characters in a text) and to transform these raw features to high level features (such as words and sentences) and predict based on these higher level features.

Key Differences Between Machine Learning and Deep Learning

Machine Learning systems rely on feature engineering which is limited to the knowledge of the expert who has to handcraft the features for detection. Machine learning-based solutions are still producing low detection rates and high false-positive rates. With Deep Learning, the algorithm analyzes all the raw data in a file, and is not limited by an expert’s capabilities. This represents a quantum leap in computer science. For future health states, this enables a more advanced level of prediction; with higher detection rates, lower false-positive rates and the ability to detect health risks effectively in zero-time.

Other differences include:

Talent: State-of-the-art deep learning solutions require deep learning experts, which are in short supply. At Marpai Labs, we have over 20 working side by side with healthcare providers in a dedicated R&D setting.

Platform: Developing a deep learning framework is an extremely complex task, which only a few companies have successfully accomplished. Publicly available deep learning frameworks are sufficient to meet the needs for most computer vision applications, but are inefficient for other applications, such as health state predictions. Very few companies are currently practicing deep learning in this domain, due to the number of significant challenges (e.g. scanning thousands of files per second).

Raw Data: Deep Learning is typically applied directly to raw data. Any answer that involves “feature extraction” or “manual preprocessing” suggests that machine learning is used. Unlike traditional Machine Learning methods where applied features must be identified by an expert and then hand coded per domain and data type, Deep Learning is applied directly to raw data without any required domain knowledge. In Deep Learning, the features are identified by the algorithm itself. The process in which the algorithms are learning higher level representation of features in deeper layers is called representation learning.

Low Level of False Positives: As the Deep Learning algorithm analyzes 100% of the data and is not subject to human error, false positives are dramatically diminished.

Brain Training Time: Deep Learning algorithms take significantly more time to train than those of traditional machine learning. Traditional machine learning algorithms typically take from a few seconds to a few hours to train, while deep learning algorithms take just several hours to train.

File Coverage: Traditional Machine Learning algorithms require different human engineered features for every file type (PDF, DOC, EXE, etc.). In contrast. Deep Learning is input agnostic, and therefore not file type dependent. This allows deep learning to be easily applied without requiring substantial modifications or adaptations.

Machine Learning Deep Learning
Domain Expert Featuring Engineering & Extraction
Requires a human domain expert to define and engineer features for conducting classification.
Looks at all the raw data in a fully autonomous manner.
Extent of Analysis A Fraction of Available Data is Analyzed
By Converting the data into small vector of features, e.g. statistical correlations, it is inevitably ignoring most of the data.
Processes 100% of Available Raw Data
One of the major strengths of deep learning is the massive number of characteristics from the raw data that is processes to obtain a decision.
Scalability Limited in its Scalability
Although machine learning can scale across diverse datasets, there is an information threshold, which if reached, additional data training doesn’t provide any further accuracy.
Improves With More Exposure
The deep neural network continually improves as the training data set constantly grows, it is the only method that benefits from scaling into hundreds of millions of training samples.

Building the Neural Network

neural network diagram

Traditional Neural Network

neural network diagram

Deep Neural Network

~90% size reduction during training

Data scientists prepare data samples to train the “brain” or deep neural network. During the “training” phase, the deep neural network is exposed to all the available raw data in a file from which it learns to instinctively identify health risks. This process takes place within 24-48 hours. Once the network has reached the prediction stage, it can quickly and efficiently predict where a health risk exists or not. The input agnostic algorithm can apply this knowledge to any sort of file. Next, the brain is compressed into a lightweight agent where Terabytes of insights are turned into Megabytes of instincts. The agent is always working to detect health risks.

Leading Deep Learning Experts

Marpai is co-founded by Dr. Eli David (PhD), a leading deep learning expert, who has published over 50 papers, and has a successful track record in building deep learning-based companies. The core team is comprised of experienced deep learning researchers and developers. Eli David has dedicated his career to pursuing the uncharted territory of deep learning to develop a product that can predict future health states, take cost excesses out of claims processing and empower proactive health by individuals. Through venturing into the depths of deep learning, his team has developed an artificial neural network brain that holds an instinctive ability to identify near-term health risks. He co-founded Marpai as the world’s first and only deep learning health plan administration system for self-funded health plans.

“Using the power of deep learning, Marpai is reshaping the healthcare mindset from reactive to preventive.”

– Dr. Eli David,  Marpai Co-founder and Chief Science Advisor

Our SMART Technology

By utilizing deep learning, Marpai achieves the dual goals of lowering the cost of claims for self-insured companies and maintaining excellent healthcare outcomes for members. We generate and collect troves of data from all aspects of our business, including healthcare claims, member engagement, internal operations, and customer experience. We systematically apply intelligent automation and AI to delight our end users, improve healthcare outcomes, and lower costs.

We use artificial intelligence in virtually every part of our business, including within our core systems as a third party administrator. These components include:

  • Claims management that enables accurate auto-adjudication of most claims, which reduces our operating costs and increases our operating efficiency. This also includes detecting and flagging suspicious claims before final adjudication.
  • Member engagement and self-service options via an intuitive and easy-to-use application, which also lowers the cost to serve employers and allows members to interact with us as often as they wish. We present the best providers in terms of quality and costs in a manner that is centered around the member’s needs and geography.
  • Care management with high-impact prediction via new, innovative AI modules, which helps us manage the overall cost of claims for our clients.
  • Continuous provider quality tracking enables us to identify the best providers as well as their usual and customary practices and related costs.
  • Smart plan design including cost optimization, performance monitoring, and risk prediction.

Our Data Power

We are creating a robust deep learning infrastructure that will allow us to analyze nearly any type of data. More importantly, this infrastructure is expected to have the capability to automatically train on multiple data types at once and obtain unified predictions from that data. Specifically, we expect our deep learning capabilities to be used for the following data types:

  • Structured data: This includes any form of data that results in a tabular database. Examples of such data include laboratory tests, claims data, and payment details to healthcare providers. Our deep learning models for structured data, which rely on fully connected neural networks, allow us to automatically train on any such structured databases and incorporate those insights into predictive models.
  • Images: While traditional image processing requires cumbersome development for each specific use case, deep learning-based models can directly operate on any image data without having to go through a preprocessing feature extraction phase. We are developing a series of deep learning models based on convolutional neural networks that could be extended to process medical images, including radiology imaging such as MRIs, CT scans, x-rays, ultrasounds, and PET scans, as well as other types of medical images, such as those used in dermatology, pathology, and ophthalmology.
  • Text: Traditionally, analyzing text data requires lengthy natural language processing, or “NLP.” Recent breakthroughs in deep learning, especially the advent of transformer networks (e.g., BERT), allow for the end-to-end training of language models, which results in significantly higher accuracy than previous NLP methods. Our infrastructure includes support for these deep learning-based language models, allowing us to automatically process textual data as well. A prime example of textual medical data is the plain text description of doctors’ notes, which include information about symptoms, diagnosis, and treatment. Even though some of this data, such as disease codes, appears in structured databases as well, the textual data contains a large amount of information that does not appear elsewhere.
  • Multi-type data: Deep learning can also be applied to multiple data sources and data types. During the first stage, the relevant deep learning models are applied to each data type (e.g., CNN for images, Transformers for text, and so on), and then, during the second stage, the processed information from those various data sources is fed into a secondary deep learning model that provides unified predictions. This is one of the major advantages of deep learning over traditional AI, as it allows the incorporation of multiple big data sources into a single unified prediction model. This approach far exceeds the accuracy rate achieved by traditional methods.
  • Data fusion: With deep learning, we have the ability to study different data types together, such as data in the form of images (e.g., CT scans) with data in tabular form (e.g., figures from healthcare claims). Putting these different data types together is referred to as data fusion.

Our SMART Health Plan Services System

Our SMART system is fully integrated with our user platform and can automatically deliver alerts to internal staff, members, and other stakeholders. It includes:

  • Mass Scale Data Fusion: Our system fuses massive amounts of data from various sources, and in many formats, into a single environment for analysis.
  • Data: Our system includes a unified healthcare schema that can ingest and make use of any healthcare data type including claims, social determinants of health, psychographic data, blood test results, pathology images, and radiology images.
  • A data lake: Our system includes a massive data lake with our unified schema that stores and structures healthcare data in order to enable analysis.
  • AI Models: Our system automatically applies our AI models to healthcare data to enable pattern recognition, advanced querying and anomaly detection.

Deep Learning for the Self-Insured Health Plan

With Deep Learning, we make accurate predictions about costly healthcare events using data types that are hard or impossible to understand with traditional artificial intelligence. This allows us to help members prevent costly healthcare events while bending the cost curve for employers.

Our deep learning technology can process, analyze, and store myriad types of data on a large scale. Our AI models recognize patterns and detect anomalies, giving us greater insights into a patient’s medical trajectory. Simulating that trajectory allows us to intervene early with the best healthcare recommendations.

With Deep Learning, we can process diverse data sources including, but not limited to, the following:

  • Medical and Rx claims data
  • Social determinants of health and psychographic data
  • Blood tests and other laboratory results
  • Pathology images
  • Advanced images from X-rays, magnetic resonance imaging (MRI), and computed tomography (CT)
  • Electronic health records
  • Health risk assessments
  • Personal health devices like SMART watches
  • Connected health devices in the home

Our technology helps self-insured employers manage their employees’ healthcare needs proactively. The key components of this technology include:

  • Deep Member Profiles: We map a member’s future health state by processing massive amounts of data. This data is fed through various deep learning modules and automatically processed.
  • Claims Trajectories: Our AI models have studied millions of claims and mapped out expected trajectories for members. The data includes structured data, such as demographic information, historical claims, and laboratory test data when available. In the future,  data will include unstructured data, such as medical images, text-based assessments, and other types of health records. Data is fed first into Deep Learning models for each data type, and then into a unified Deep Learning model that finds patterns and insights for the member.

Deep Learning models are especially superior in finding non-linear patterns and correlations. For example, a new data point X — that seems to be completely unrelated to prediction Y — may actually affect it through complex non-linear patterns that are very difficult for humans and traditional artificial intelligence to find. Deep Learning models are very good at finding those patterns.

  • Alerts: When our system detects a near term health risk, we send a proactive health alert and update the entire model. These alerts direct the Care Guide team to reach out to an at-risk member.

For example, we may send an alert to a member who is likely to develop Type 2 Diabetes in the next twelve months. Our team would reach out to explain the prediction and guide the member to the appropriate provider.

Our AI-power predictions help members get in front of costly events and prevent them. This averts suffering, lowers claims costs and improves the ROI of the plan.

Mailing List