The integration of medical imaging, artificial intelligence and radiomics is set to play a pivotal role in precision and personalised healthcare – and we are delighted to be backing a pioneer, Quibim, and its founders, Dr. Ángel Alberich-Bayarri and Prof. Luis Marti-Bonmati.

Here is the story.

Every year nine million people die from cancer alone, all too often because they are diagnosed late. Being able to discover an anomaly early is key to recovery. Preventative measures can even delay disease onset altogether. Detecting subtle changes in physiology and pathology before they can be detected clinically and doing so in a non-invasive and patient-friendly way is a prerequisite to adoption.

Biomarkers are the weapon of choice. They can be anything from DNA, RNA, protein, metabolites measured from samples such as blood, tumour material, urine or saliva to imaging such as digital pathology and radiology. Conceptually, they have been part of medicine for centuries; body temperature and heart rate are among the oldest.

Beyond diagnosis, biomarkers also help us better understand the mechanism of action of drug candidates and guide therapy, monitor recovery or relapse, and change the way research is carried out. Patients are no longer benchmarked against each other; rather clinicians assess the evolution of a marker and individuals act as their own control.

Sometimes biomarkers become vital early in life, as in the case of paediatric cancers – one of the therapeutic areas where Quibim excels. The company is leading Primage, a consortium of 16 European partners dealing with medical imaging, artificial intelligence and cancer treatment in children.

Sometimes they suddenly become critical to us all, as with a pandemic. In March, Quibim became the official European platform for high-throughput screening of COVID-19. Shortly after, the Radiological Society of North America (RSNA), with more than 52,000 members from 153 countries, joined the initiative with the common goal of creating a secured global medical repository of COVID-19 cases.

Mostly, biomarkers are companions for life. Their use helps to improve diagnostic accuracy, predict prognosis or response to treatment. Quibim offers more than 20 applications for a range of conditions including numerous cancers, Alzheimer’s, osteoarthritis and liver disease. Its AI platform extracts and quantifies disease-specific biomarkers from medical images with ultra-high accuracy and is already used in over 70 hospitals and 11 clinical trials across the world, with millions of images correlated with medical records processed to date.

When They Go Deep, We Go Wide

Quibim is on a mission to build the reference radiomics platform for whole-body solutions. With COVID-19 for instance, whole-body imaging allows for the discovery of individualised phenotypic signatures of a virus that affects multiple organs and each patient in unique ways. The strategy is rather contrarian as the industry trend is instead to take one organ or one imaging modality, and go deep – consequently the space of medical imaging analysis is littered with one-trick ponies.

We believe that while specialism may be instinctively attractive, it simply does not scale in the context of radiology where clinicians are also looking to simplify and optimise workflows. Not a surprise then that many of their clients are every so often first-time users of radiomic solutions. It is not a lack of awareness or interest that has led them to turn down other solutions, rather the inherent narrowness and rigidity of specialised clinical offerings or the incompatibility with their own workflows.

Quibim was built by radiologists for radiologists and its founders are at the forefront of the field. For instance, they established the step-by-step process for the development of imaging biomarkers, eventually adopted in 2013 by the European Society of Radiology (ESR) as the industry guideline. One of the first techniques they developed as early as 1994, soberly dubbed “temporal reconstruction images”, is now prevalent and known as Dynamic Contrast Enhanced MRI. It led to the discovery of permeability coefficients such as Ktrans used in oncology to analyse tumour angiogenesis. This imaging biomarker has since been shown to be related to vascular endothelial growth factor in tissues.

Quibim is built on this legacy, continues to publish extensively and has already assembled one of the richest catalogues of biomarkers and non-invasive detection methodologies in the world.

Visualize. Annotate. Quantitate. Discover. Repeat.

A quick PubMed search reveals nearly one million citations associated with biomarkers. The hunt for new ones to drive precision medicine is at an all-time high yet validating them is a challenge and only a few dozen have so far made it through the FDA.

First, to be clinically meaningful, they must be biologically accurate and disease relevant. Measuring biomarkers provides valuable information on certain metabolic processes: the more specifically a biomarker is connected with a disease process, the more precise the information that can be derived. Establishing this “ground truth” is as fundamental as it is difficult. In some cases, they are only a proxy, sometimes a notably imperfect one – think of PSA for prostate size.

Then, a biomarker faces two more adoption hurdles, this time technical: sourcing on the one hand, and detection and/or quantitation (aka. technical accuracy) on the other. Sourcing because current sampling methods are currently invasive (blood tests, colonoscopy, tissue biopsy, etc.) and hence not used regularly enough, and more importantly, not early enough. And technical because, in the field of biomarkers, the ability to demonstrate accuracy is more challenging than in most analytical disciplines as the analytes are endogenous, heterogeneous and often structurally different from the calibrator. It is not uncommon that different methods produce different results using the same sample.

This is where Quibim comes in – covering all three fronts. Its founders built their careers establishing clinical ground truth. Its radiomics platform leverages precision and non-invasive modalities used routinely in clinical set-ups. And finally, its quantitative imaging biomarkers (QIBs) are designed to deliver ultra-high accuracy in measurements, tackling a fundamental shortcoming of radiology – subjectivity.

Productivity Beast

Imaging is becoming central to multi-parametric healthcare assessment. The development of predictive models, taking into account all types of clinical, pathology, molecular and imaging information to forecast valid disease-related outcomes, is a hot topic.

For example, blood biomarkers such as circulating tumour DNA, while becoming sensitive enough to detect the presence of a disease, are rarely adapted to easily localise or stage tumours. This is where imaging reigns. It also offers a fuller picture to clinicians as it can capture heterogeneity across the entire volume of tissue (for instance, looking at a tumour or an organ in its totality), allowing for finer clinical decisions. It also allows for investigations where no other techniques can – neurology, for instance, through virtual brain biopsies, critical to the early detection of conditions such as Alzheimer’s, Parkinson’s or glioblastoma.

Its no surprise then that the demand for medical imaging services is fast increasing, outpacing the supply of qualified radiologists and stretching them to produce more output, without compromising patient care. Healthcare systems face a dual problem; not only are we lacking in personnel, but radiology also covers many subspecialties and clinical pathways and few departments have the necessary skillsets. For instance, data indicates that on average as much as two thirds of chest X-rays are not analysed due to the lack of manpower. Or, less than 5% of radiologists are trained to analyse MRI data for prostate cancer.

Yet, in the UK, the NHS spends around £2 billion per annum delivering imaging services in public hospitals. It is crippled with operational inefficiencies; in the last few years, the NHS had to settle legal claims in excess of £300m made against imaging departments in England due to missed findings or delays.

The AI-modulated tools developed by Quibim help radiology departments to tackle many of these challenges, transforming their efficiency and accuracy. Their algorithms run in the background, monitoring in real-time the output of MRIs, PET and CT scans and other modalities installed, flagging and quantifying pathological cases to be prioritised. Their platform outperforms across disease areas, auto-generates reports, triage cases and improves overall turnaround times. Tedious and repetitive tasks are also taken over and the resulting reduction in workloads allows radiologists to focus on complex cases. It is a comprehensive solution that consistently displaces other providers.

The early bird may get the worm, but the second mouse also gets the cheese.