Different patients with the same type of cancer can respond very differently to a specific treatment, and as a result, outcomes vary greatly between patients. Many clinical, pathological, and genetic factors make the diagnosis and choosing the right treatment for an individual patient increasingly complex.
In Medical Delta Cancer Diagnostics 3.0, the newest imaging techniques will be used together with machine learning to provide diagnosis faster and to better predict the course of the disease.
Creating non-invasive methods
At the moment, biopsies are taken to obtain information about the genetic and molecular characteristics of the tumor. Taking a biopsy is an invasive method that can sometimes be risky. Researchers from the Medical Delta Cancer Diagnostics 3.0 program believe analyzing tumors could be improved by extracting more information from MRI images.
Advanced MRI techniques
Within this program, which initially focusses on brain tumors, advanced MRI techniques will be used to gather relevant information about the tumor. The researchers will do this using standard MRI scans, as well as the newest MRI techniques and hardware. “For example, we will use a 7 Tesla MRI, which has a higher magnetic field and provides images with a better resolution and more information,” says Thijs van Osch, professor in the radiology department at LUMC.
The aim of comparing information gathered through biopsies and information visible on the MRI scans is to deduce the genetic and molecular characteristics of the tumor from the MRI-data: a ‘virtual biopsy’.
In addition to obtaining better MRI images, this Medical Delta research program links them to machine learning to unlock all the information hidden in the different MRI-contrasts. Ultimately, this is expected to contribute to an even better treatment choice tailored to the individual patient, and to enable better monitoring of the tumor during treatment.
“It would be of enormous added value if we were able to better predict on the basis of all those MRI characteristics how the tumor will evolve and what that means for the patient in the future,” says Johan Koekkoek, neuro-oncologist in LUMC and MC Haaglanden.
For doctors it would very helpful to get guidance in making the right diagnosis and selecting the best treatment. “Currently, diagnosis is based on a visual assessment of the MRI scan by a radiologist,” says Marion Smits, professor of Neuroradiology at Erasmus MC. “Machine learning techniques will help us to include a lot more information from the MRI scan in our assessment, and to make such diagnoses more objective.”
Machine learning solutions are also in development in the field of digital pathology of biopsy material. Sjoerd Stallinga, professor of Computational Imaging at TU Delft: “Analysis and classification of tissue morphology is a very suitable task for modern deep learning methods.”
The close collaboration that has been established between Erasmus MC, TU Delft and LUMC with MRI experts, computational scientists, engineers and clinicians is a key success factor for realizing the ambition to enable biopsy-free image-based diagnosis of cancers. The support to provide a personalized treatment selection and treatment monitoring is envisioned as a template for similar Medical Delta innovations in the care of other cancer types.