Vendor agnostic, fully query-able, extensible

The Quantivly unified data layer

You can retire your notebook and manual logs. Quantivly cleans, harmonizes and aggregates various data sources on-the-fly (e.g., image data, image metadata, and radiology information system (RIS) messages) to build a unified, unbiased, objective source of knowledge about scanner operations.

Our unified data layer is like a middle layer technology infrastructure that abstracts away machine/vendor complexity so that you can seamlessly ask questions, regardless of the scanner vendor.

The many data siloes in medical imaging

The digitalization era has created data siloes that serve different verticals, e.g. the radiology information systems (RIS) and the picture archiving and communications systems (PACS). Within each system, various vendors have implemented their own flavors of each standard. The result is a situation where the data exists but cannot be synthesized to form a complete picture of operations.

Beyond data, metadata has been substantially overlooked. Yet, it is essential to understand the context of data. We believe that many analyses and AI applications will increasingly need to incorporate context to content to succeed, but it is critcally missing today.

"Data is content, and metadata is context"
– Data and Goliath

The Quantivly harmonization engine
(or how to solve this data liquidity problem)

We built a software platform that cleans, harmonizes, and aggregates various data sources on-the-fly to build a unified, unbiased, objective source of knowledge of scanner operations – we call it the Harmonization Engine. We first focused on the data created by imaging equipment (DICOM data and metadata) and scheduling systems (HL7). But we have many more to come.

Our key philosophical guiding principle: we take a “bottom-up” approach – this means focusing on data harmonization, augmenting our ontology with critically important concepts, and then building many applications on top of the data layer, as opposed to starting with an application and making the data fit it.

And we worked really hard on optimizations. The result: blazing fast queries. You don’t need to classic ETL (extract, transform, load) – you can just query the database directly.

Key characteristics

Harmonize DICOM

While DICOM provides a basic standard for describing the geometry of images for radiological interpretation, it fails at providing an ontology that describes imaging exams – it was designed for a different purpose and has been incompatibly extended by different vendors. After years of R&D, Quantivly has built an ontology that starts from DICOM and builds upon it to construct the first unified, vendor-agnostic description of imaging exams.

For example, we extract the concepts of imaging “exams” and “acquisitions” from DICOM studies and series, critical to accurately understand procedures’ duration and scanner utilization. We also extract the concept of volume, critical to accurately capture imaging parameters (e.g., that multi-echo TE MR sequence, with multiple TE in the same DICOM series).

Harmonize HL7

DICOM provides information about what was acquired, when and how; but this is only half the story. Quantivly’s data layer unifies information from the radiology information system (RIS) via HL7 to compare “what happened” with “what was scheduled”, allowing for a continuous feedback loop to improve imaging operations.

An example? You can finally deep dive into delays; compare scheduled slot size and exam durations; and add granularity to your data (e.g., in/out patient, sedation, etc.).