Answers Not Analytics™
Our proprietary algorithm has the potential to identify important insights and provide actionable answers based on your data.
CureMetrix is developing an algorithm that would allow physicians to UNDERSTAND the factors that drive each outcome.
The CureMetrix image analysis platform brings transparency to all aspects of the data – elevating visibility and enabling real-time, data-driven decisions. Unlike a neural network that is a ‘black box,’ the CureMetrix algorithm output is a “clear box” model that displays the specific factors that lead to an outcome.
We seek to provide radiologists with advanced technology for mammography that allows visibility and understanding of the factors driving outcomes. We hope this will allow physicians to focus on the most relevant data and discard irrelevant variables to help you converge on answers more quickly and accelerate results.
Using smart sampling techniques, CureMetrix breaks through the analysis paralysis of big data by honing in on the most relevant information. Searching through and analyzing vast datasets can be expensive, inefficient and ultimately futile. With big data it is possible to look at variables or attributes that have no impact on outcome. These factors combined can hinder true, data-driven decision-making.
The CureMetrix approach to big data strikes the perfect balance of data volume and representative samples without losing fidelity within the data. We are developing our learning algorithm to explore critical image data, identify what attributes are relevant, discard those that aren’t, and create samples that allow us to quickly converge on answers physicians need.
The CureMetrix image analysis platform has the potential to produce highly accurate, predictive models from a wide variety of data, including static data that is unlikely to change (such as patient demographic information) or data collected over time (including patient vitals, or sensor or image data), as well as historical or real-time streaming data. We elevate visibility and offer insights that have an immediate, cross-functional impact.
We’re nimble and work effectively with real-world data that is often extremely noisy, multi-variant, and changes over time. Furthermore, CureMetrix has a unique ability to combine different data types into an empowering and revealing dataset for analysis.
The CureMetrix image analysis platform is easy to deploy since it runs on any typical desktop, laptop, cloud or mobile device. It is so lightweight that it can quickly run in-memory calculations that fit into a single cell of a spreadsheet.
In predictive modeling, it is important to operationalize the answers we find in ways that improve outcomes. This means that our models need to deliver answers quickly. Too much complexity in the model can result in over-fit (when the model is overly customized to specific data), whereas not enough complexity results in under-fit (a model that is too simplistic). The mathematical foundations of the CureMetrix approach are designed to balance the competing risks of under-fit and over-fit to identify the level of model complexity that guarantees the best predictive performance.
As data changes, so does the model driving outcomes. As we continue processing novel instances of physicians’ data, our algorithm learns and becomes more accurate. The model adapts easily as changes occur, ensuring highly accurate answers at an optimal, lightweight size.