Select Region:
Select Language:

The medical world nowadays is facing an increase in the demand for medical examinations and a shortage of radiologists to cover that demand. There are also very limited and complex solutions to assist the radiological workflow and a decrease in response capacity. In this context, AI has emerged as a promising technology playing a huge role in finding solutions to these problems and enhancing the standard of patient care. 

“Artificial Intelligence in Radiology” was the main topic of the last week’s webinar hosted by the Royal Academy of Medicine and Surgery of the Region of Murcia, which was presented by different experts in the field. Among those, Ana Jimenez Pastor, Senior Data Scientist & AI Innovation Lead at Quibim, was in charge of taking the audience through the main different applications of AI in radiology. She particularly emphasized how Quibim can significantly improve the processes and workflow of radiologists and clinicians through other AI-based solutions. 

Figure 1*

Quibim AI-based tools intend to accelerate and automate laborious tasks, as, more than ever, physicians need solutions that can help them work faster, better, and with the highest accuracy.

  • Virtual Liver Biopsy to help diagnose diffuse liver diseases: To quantify the concentration of fat and iron in the liver, Quibim uses MRI exams to extract the same information we can get from a histopathological biopsy but avoid its main disadvantages, such as its invasiveness nature, sample heterogeneity, and intra-/inter-observer variability. To identify the liver region within the MRI exam, Quibim uses a specific AI architecture,  Convolutional Neural Networks (CNN), that automatically segments the liver, accelerating the entire pipeline of Fat Proton Density (PDFF) and R2* quantification.

Read more about this: Precise whole liver automatic segmentation and quantification of PDFF and R2* on MR images.

Liver Fat and Iron Quantification

Figure 2*
  • Prostate lesion detection: MRI allows the creation of automatic tools to perform a “virtual biopsy” of the prostate. We propose a new algorithm to extract findings using non-supervised clustering techniques applied in the T2, DWI (and T2, DWI, DCE if the study is not bi-parametric but multi-parametric). In addition, these multi-parametric or bi-parametric acquisition protocols allow the quantification of the Apparent Diffusion Coefficient (ADC) using the DWI, and/or to perform pharmacokinetic modeling using the Dynamic Contrast-Enhanced (DCE) sequence. Therefore, using our AI model, a clusterization-based algorithm, we can automatically segment the whole prostate gland and highlight potential tumoral areas or certain patterns related to prostate cancer.

Prostate Nosolgic Maps

Multi-Parametric MRI

Figure 3*

The tools developed by Quibim are a down-to-earth example of AI-based solutions, by assisting clinicians in converting the qualitative data in the medical images into quantitative data (imaging biomarkers), making it easier and faster for them to detect anomalies enhancing patient care.

However, our AI applications are not only applicable to radiology, but also to biopharma companies. Quibim solutions can help biopharma companies to develop treatments and drugs more efficiently. In fact, the biopharma industry relies more and more on medical images and is increasingly dictated and regulated by evidence-based guidelines.

In summary, AI tools can help to reduce radiologists’ workload by the automation of some tasks and expedite biopharma drug discovery pipelines for fast and cost-efficient results. However, we cannot forget that the development of AI models is not a straightforward process. Thus, an understanding of the scope and the ground truth available to follow the appropriate validation process and to guarantee its generalization and reproducibility in new clinical settings is required. Additionally, a deep knowledge of the unmet needs in daily clinical practice is essential, ensuring synergies between research and product development. Ultimately, these structured development processes will allow the creation of imaging biomarkers panels to meet those clinical needs.

Watch the entire webinar to learn more about the applications of AI in clinical practice.


Figure 2. Jimenez-Pastor et al. (2021) Precise whole liver automatic segmentation and quantification of PDFF and R2* on MR images. European Radiology.

Figure 3. Hanahan, D., & Weinberg, R. A. (2011). Review Hallmarks of Cancer : The Next Generation. Cell, 144(5), 646–674.

Katherine Wilisch Ramírez