.png)
Introduction
Radiomics is a Belgium-based cutting-edge MedTech scale-up. Their team uses artificial intelligence on medical images to deliver insight-based decision-making to optimize pharmaceutical and biotech companies’ clinical trials in oncology.
The Radiomics vision is a simple yet laudable one, that RedBrick AI also shares; “make precision medicine a reality via AI & image analysis”.
We recently had a chance to sit down with Medical Communication Manager, Fabio Bottari to discuss the technology and products developed by Radiomics , as well as how their team uses RedBrick AI’s platform.
What does “radiomics” mean?
In medicine, radiomics is the discipline of extracting quantitative features from medical images to uncover information invisible to the naked eye. After the features are extracted, they can be used to build statistical and machine learning models for descriptive and predictive analysis.
Why is that important?
Say an unfortunate patient has been diagnosed with non-small cell lung cancer.
To better treat this patient and assure the highest chance of survival, the current standard criteria sometimes fall short. For this reason, Radiomics want to quantify all the information inside medical images, looking at all the lesions and based on objective and reproducible parameters.
Radiomics can extract what the human eye is physically unable to interpret from the images and capture details about things such as a tumor's shape, size, texture, intensity, and much more.
These descriptors can ultimately be used to predict outcomes such as expected tumor growth and potential response to treatment. But also, to describe the type of tumor and the inherent biology of the lesion (specific mutations, micro-environment, protein expression, etc.).
What does the team at Radiomics.bio do?
Broadly speaking, Radiomics builds artificial intelligence to analyze medical images and guide clinical trials.
But that’s perhaps a bit too broad of an answer, so let’s break things down a bit.
In the pharmaceutical industry, clinical trials are a crucial step in the process of getting a new drug approved. These trials are typically broken up into four phases, where the early phases (Phase I) is focused on assessing a drug’s safety and appropriate dose levels, and later stages (Phases II to IV) are focused on assessing the efficacy on patient populations and long-term effects.
Clinical trials can be a massive and expensive undertaking, as they can last for several years. On top of that, drug manufacturers face many challenges that can add to the risk and cost of clinical trials, including recruiting the right participants for the later phases and executing the trials with the correct dosage.
As the team at Radiomics has observed, pharmaceutical organizations often resort to trial and error to identify a drug's maximum and optimal dose during the early stages of a trial.
In contrast, Radiomics ’s approach can identify which dosage of a drug will be most effective by tracking how a patient responds to treatment from the very beginning of a trial.
This type of information is immensely useful to pharmaceutical organizations and contributes significantly to their ability to better plan later-stage trials.
In the later stages of a trial, Radiomics helps organizations identify which patient population is most appropriate for a drug.
As Fabio said:
Where RedBrick AI Comes In
The basis for Radiomics technology is the extraction of radiomics features from medical images. To do so, the CT or MRI scans needs to be segmented (i.e. identified and delineated) and annotated by experts, to select the region of the body (organ or lesion) from which the radiomics features will be extracted. So the first step of each radiomics analysis is the creation of a large annotated dataset.
In the past, the team at Radiomics used locally-installed open source software combined with in-house tools for annotation and quality control.
As Fabio describes it:
Now, Radiomics uses RedBrick AI’s medical data annotation platform to prepare ground truth data for training their segmentation algorithms and for the extraction of radiomics features.
RedBrick AI also offers a wide variety of tools to effectively manage projects & workflows, as well as APIs to integrate with data stores and Machine Learning Operations. This tool suite includes comprehensive quality control features, including the ability to create multi-stage review workflows, measure inter-annotator agreement, and track individual labeler quality scores.
The feature that Radiomics uses and find most useful is the review workflow.
Everything we build at RedBrick AI, from the supported data formats (DICOM, NIfTI, etc.) to the viewing tools (multi-series, 3D, MPR, etc.) to the annotation tools (intelligent contouring, threshold-based segmentation, etc.), is built for radiology data and the people who work with it every day.
At RedBrick AI, we firmly believe that technology can make the world a better, healthier place. The work of startups such as Radiomics is both inspiring to behold and serves a key role in the advancement of medical technologies.
We’re thrilled to continue our collaboration with Radiomics and have the chance to watch as they transform clinical trials through insight-based decision-making.
