Central Imaging Lab
Medical images are used in many therapeutic areas and contain a goldmine of information. Part of this information is traditionally extracted by radiologists and with the advances made in image processing and artificial intelligence, computers are now able to go deeper in the extraction of almost every detail which is hidden in the image. In clinical trials, centralized and automated analysis of medical images allow the removal of human-induced variability and bias. Thanks to our extensive experience in handling medical imaging variability, applying AI and trial statistics, and working with pharma companies, our goal is to position Radiomics as the next generation imaging CRO. For this, Radiomics provides a full package of services.
QC
Quality of the medical imaging scans is fundamental for the success of the radiomics analysis and its subsequent interpretation. Radiomicsworks with international medical and clinical data standards such as CDISC (Clinical Data Interchange Standards Consortium) and DICOM (Digital Imaging and Communications in Medicine).
CDISC is an organisation developing standards used for the different steps of a clinical trial including analyse and submission.
DICOM is the international standard used for the medical images, including transfer and storage.
This allows us to respect FAIR principles (Findable, Available, Interoperable, Reusable) regarding data.
Moreover, Radiomics helps its clients with the:
- Set-up of clinical trials
- Providing optimal image acquisition protocol for high quality data
- Imaging analysis phase:
- Scans are free of confound information such as artifacts
- Scans contain the correct region of interest (ROI)
- Scans are acquired following appropriate acquisition protocols.
Radiomics bases itself on years of experience in quantifying variabilities in images.
If problems are identified, our team of expert radiologists and AI scientist will guide clients through the available mitigation strategies.

Segmentation
The first step in the imaging analysis process is the delineation or segmentation of the region of interest (ROI). At Radiomics this action is performed both manually or in a fully automated manner on CT, MRI, PET and Rx images. All our manual segmentations are carried out by a team of trained staff members and the results are always validated by at least one board-certified radiologist.
Additionally, based on our ever-growing dataset of available and validated segmentations, our R&D team continuously builds automatic segmentation tools. These solutions can replace the manual segmentation once their accuracy is demonstrated to be non-inferior to a set of board-certified radiologists.
Currently we have developed automatic segmentation tools for the following:
Additionally, based on our ever-growing dataset of available and validated segmentations, our R&D team continuously builds automatic segmentation tools. These solutions can replace the manual segmentation once their accuracy is demonstrated to be non-inferior to a set of board-certified radiologists.
Currently we have developed automatic segmentation tools for the following:
And our R&D team is currently working on automatic segmentation tools for airways, pulmonary vasculature, vertebral column, gastric cancer, prostate cancer, brain cancer, breast cancer and head & neck cancer.

Radiomics
Our founders are the inventors of the word and concept of radiomics. Radiomics is a synonym for advanced image quantification and this is at the core of what we do at Radiomics. Based on the segmentation of the relevant region of interest, the radiomics analysis quantifies all the information contained in the medical images.
This quantification can be based on handcrafted or deep learning features. Handcrafted features are mathematical equations representing the size, shape, texture, and intensity of the region of interest. Deep learning features use neural networks to identify relevant characteristics of the ROI that cannot be defined upfront.
These features can be used as endpoint linked to the efficacy and safety of treatment and can be combined to predict disease presence, progression and therapeutic outcome.

Interpretation & Reporting
At Radiomics we know that data alone is not enough to lead to concrete solutions addressing the unmet medical needs of our clients. The high-resolution dataset generated by our radiomics analysis requires adapted trial statistics approaches. Therefore, our team of experienced data scientists and statisticians is specialized in guiding our clients during the trial data analysis steps, providing data interpretation, and reporting at different levels for scientific, executive, and market-oriented audiences. This will guarantee that data is used and presented in optimal ways, to make sure that our conclusions are insightful but not overly optimistic.
