Info

TraIT welcomes its 3000th user

Interview with Alberto Traverso

The Translational Research IT (TraIT) project is happy to welcome its 3000th user.

Alberto Traverso, researcher at Maastricht University and MAASTRO Clinic, recently applied to use the TraIT infrastructure for his study on lung cancer imaging. Traverso will use the National Biomedical Imaging Archive, one of the TraIT tools, to establish a centralized repository for imaging data.

The Medical Physics doctorate comes from Italy originally and moved to Maastricht in October, to work in the Knowledge Engineering group of Prof. André Dekker. His research focuses on management of clinical data and developing tools for clinicians to take decisions in relation to cancer therapy.  “In our group we work a lot on radiomics, which is quite a new field. It’s a way of analyzing medical imaging data, especially scans from patients in the hospital.”

Roberto Traverso

Radiomics

Within the big “Radiomics” project, Traverso and his colleagues are developing tools to automatically analyze the imaging data. “Now doctors look at the image, and of course they can say something about the properties of the tumors, but there is a lot of information invisible for the human eye. Therefore, we are working on an automatic way of extracting this information and relate it to clinical applications”.  Next to that, the group is working on data management and the quality of data. “We are building a platform for radiomics, an infrastructure to automatically collect all data. This is an international effort, with hospitals around the world involved.”

 

Using the National Biomedical Imaging Archive

The idea of the study was to have a common tool where all partners of the project can submit and collect anonymized data. “In principle you have quite a lot of data, because every scan that is made in the hospital can potentially be used. But there are two problems. Firstly, the data are decentralized, so it is difficult to get access. Secondly, data are often very different because the protocol of storage is different. We need to define data in a universal way, making it readable for machines.” The solution is the National Biomedical Imaging Archive (NBIA), that is used as a central repository, one of the most fundamental parts of the study. “The nice part of the tool is that it includes a quality control before publishing this data. This way you create homogenous data that can be further used for research.”

 

Multi-site, multi partners

For Traverso, it is the first time working with data on such international scale. “I was used to working with local data. Therefore, there was never a need for a big central repository. My supervisor (André Dekker) already used the NBIA tool in several projects, so it was natural that we started working with it.” The radiomics project is a big project, with the aim to collect 3000 patients from 10 partners from different sites, in the US, Asia and Europe. “We started working with the tool in January. At this moment, we are training our partners to work with it. And every week there are new people wanting to take part and share their data.”

 

Towards personalized treatment

The project of Traverso and his colleagues will run for at least 4 years. The first goal is to build the radiomics platform. Eventually, retrieving more information about the properties of the tumors offers the opportunity to create a more personalized treatment. “You don’t want to give every patient the general treatment. As each tumor is different for each patient, you want to differentiate this, according to the properties of the tumor and taking into account what you see on the image.” Currently invasive biopsies are needed for diagnosis. “If you can read more from the images, we could use them as predictive tools for cancer therapy and find the treatment that works best for the specific tumor.”

 

Machine-learning

Eventually Traverso would like the data to be readable for machines, so they can be fully automatically analyzed. “When I go to conferences I notice that clinicians are wary of this automation. They fear computers will substitute their jobs. This is not the case; it is always based on human design. We can use computers to save time, so doctors can focus on their expertise. Computers can notice features that are invisible for the human eye. But we need doctors to interpret the data.”

 

Different languages

The research project combines the interests of Traverso in working with clinical data and imaging processing. His experience in Torino, especially working with prof. Piergiorgio Cerello, (Turin Section if INFN, made this possible. “I really need to thank Piergiorgio for all the effort put in growing me as young researcher”. “After I finished my PhD in Medical Physics in Torino I applied for this position at Maastro Clinics. I like the broadness of the project, working with different people with various backgrounds, from computer science to medical researchers. It’s challenging. We speak different languages, look at the problem from different viewpoints. That’s what research really is about. Learning from each other and build new common knowledge.”

 

For more information on the project, please visit the MAASTRO Knowledge Engineering pages.