16 October 2023
The aim is an ambitious one: to create a standardised system for storing patient data in the cloud, duly anonymised, so that it is accessible, in complete security, to researchers who wish to study it. And also, one day, to clinicians and patients themselves, to be consulted via the internet, anywhere in the world.
A novel infrastructure for storing, consulting and analysing medical data will be presented at the Open Day of the Champalimaud Foundation (FC) Breast Unit, which is taking place today, October 16. It was created by the Unit's Digital Surgery Lab, a multidisciplinary team led by surgeon Pedro Gouveia, in partnership with the telecommunications company Altice and the Portuguese company BMD Software.
"Today, in Portugal and in most centres in Europe, when you need access to medical image data – to do research, for instance – you have to go to the hospitals to consult the patient image archives and burn the information onto a CD," explains Gouveia. In the 21st century, we're still using physical disks to transfer information!"
This is not a secure method, Gouveia stresses, because it is not possible to control who has access to the CD data, nor to be sure that the process of anonymising it (which is essential to ensure compliance with the data protection regulation, the GDPR) has been done correctly. "Now, for the first time, we have been able to demonstrate that it is possible to transfer and store digital files securely, virtually, via the internet, using tools that already exist in hospitals," adds Gouveia.
The fact that the data can be stored in the cloud makes it much simpler and more immediate to access to accredited people. But these researchers go further: if the format of this data is properly standardised, they explain, it will be possible to envision a future in which hospitals and research centres can transfer medical data between themselves in a way that is totally transparent to the end user.
Gouveia gives an example of the current situation: today, each hospital can choose its own date format, namely North American or European. "How can we expect two systems to communicate if this is not standardised?" he asks.
"We have already managed to standardise and achieve interoperability. The Hospital de Santa Maria, for example, could get medical images and clinical data on breast cancer patients treated at the Champalimaud Foundation without having to first figure out how we format dates or other elements. Because all of this would already have been standardised."
The creation of the new infrastructure is just the first phase of a project – called MetaBreast – that Gouveia has been developing over the last few years with colleagues from the Breast Unit and other partners. The project was recently approved to receive funding from the Portuguese Resilience and Recovery Plan (PRR) – around 1.4 million euros of which will go to the CF (https://fchampalimaud.org/news/future-of-breast-cancer-surgery-is-coming).
The new infrastructure will, for the time being, be used for research (developing software, testing the reliability and security of stored data, etc.), but the team's aim is to create, in a second phase, a commercial medical device (in part funded by the PRR), which will allow doctors to access health and image data via augmented reality goggles.
Here we are entering the so-called "surgical metaverse" – which, for Gouveia, is undoubtedly the future of surgery. Because augmented reality goggles connect to the internet, the new infrastructure should make it possible to access patient data via these goggles directly in the operating room.
More precisely, MetaBreast's ultimate goal is to create a medical device that, thanks to software or a service, uses the internet to enable breast surgeons (and others later on) to access the imaging and clinical data of cancer patients in real time during surgery via augmented reality goggles. Surgeons will thus be able to superimpose various radiological images of the patient onto the patient's real body, live in the operating room, making surgery more precise – or even to directly consult other types of clinical data about that same patient if and when necessary (https://fchampalimaud.org/news/augmented-reality-technology-used-first-…).
The first step of the MetaBreast project was initially aimed at creating an online database of the so-called "pseudo-anonymised" records of several hundred breast cancer patients, with annotated medical images, photographs of the torso and other relevant data for each patient. Pseudo-anonymisation not only guarantees data security (GDPR), thanks to the assignment of a key (a secret code), to each patient's data, but also opens the possibility for the doctor who is treating a patient to identify that patient, in case of need, using that same key. "Pseudo-anonymising the data is a more ethical way of doing things, because if any relevant discoveries are made about a patient, the doctor treating that patient can act accordingly," explains Gouveia.
With the development of the infrastructure being presented today, this part of the project has now become more ambitious. "We had already talked about creating an online database," recalls Gouveia. "But what we're creating now is much more than that. An online database is normal for any project. What we've done here is something completely different: we've created an 'interoperability system' and health data storage", digitally connecting the clinical system to a cloud data repository.
That is the novelty, explains the surgeon. "Interoperability is the keyword," he says. It means no longer relying on manually extracting patient data from a CD or an Excel spreadsheet, and getting computer systems to communicate in real time, transfer data between them and anonymise or pseudo-anonymise it from end to end, automatically and securely.
Hospital medical image archives are already being stored in standardised systems, called PACS (Picture Archiving and Communication System) that were developed using a standard called DICOM (Digital Imaging and Communications in Medicine). DICOM’s have replaced conventional radiological films. In each hospital, these medical image archives receive the image files directly from the equipment that, at that hospital, performs radiological examinations such as CT scans, MRI scans or X-rays.
Non-imaging data – which at the Champalimaud Foundation, for example, is contained in an electronic medical record – will also be integrated into the new infrastructure.
"All the image data and clinical processes from the Breast Unit will be transferred to two computers on Altice's private network at their headquarters in Lisbon," says Gouveia. To do this, the team created, within the Champalimaud Clinical Centre's cloud, a so-called gateway PACS – that is, a transfer PACS – that will receive the data from the Breast Unit's clinical PACS, pseudo-anonymise it and then send it outside the clinic's cloud, to a third PACS, called the research PACS, installed on Altice's servers. This latter PACS will be available, via VPN, to the Foundation's researchers and members of the MetaBreast consortium, who will thus have access, not only to the imaging data, but also to the non-imaging data stored by Altice.
"The Breast Unit already has 3,500 patients with structured clinical data who will therefore be able to 'travel' from PACS to PACS to their final destination," adds Gouveia. And even the patients themselves, if they so wish, will one day be able to access, with the appropriate credentials, their entire electronic medical records from any hospital or clinic that has a PACS.
The gateway PACS and the research PACS are now operational and test transfers from one to the other have begun. The team intends to test the reliability of the infrastructure with several hundred patients recruited for that purpose at the CF (25% of whom have already been enrolled – that is, more than 100 patients since February 2023). "For this clinical study", explains Gouveia, "we are collecting MRI data from the patients in two positions – face down and face up – and also building maps of the surface of each patient's body, with a specific device, in a position analogous to that adopted during surgery."
The anonymised MRI scans will also allow João Santinha, a specialist in artificial intelligence applied to medical images at the Digital Surgery Lab, to train the artificial intelligence algorithms he is developing. In this way, without having to directly access the Foundation's clinical PACS, which contains the patient's identity, he will be able to automatically segment the different breast tissues and identify where the tumour is, where the fat and fibroglandular tissue are. These algorithms will then make it possible to build personalised 3D models of the patients that surgeons will one day routinely visualise, if all goes well. "To be able to train our AI algorithms, we need to transfer the data of at least 400 patients to the new infrastructure," explains Gouveia.
What's more, it may even be possible, in the future, to take advantage of data present in medical images, but which is not yet being explored, to improve patient quality of life. Santinha gives an example: "Breast cancer patients, when they have conservative surgery, always have to do radiotherapy afterwards to ensure that the disease has been eradicated. They therefore need to have a CT scan to plan their radiotherapy so that we can identify the organs at risk and the areas where we want to have certain doses of radiation. Now, there is a lot of information that is visible on these images, but which we are not using today, such as cardiovascular problems that patients already have or may come to have, risks of cardiovascular disease, osteoporosis, obstructive respiratory disease."
These signals could one day be used to identify women with breast cancer who are at risk of other diseases and to adapt treatments in order to minimise these risks or the effects of these diseases.
In the meantime, the team is preparing to create a spin-off company to market the AI and augmented reality medical device that will be developed as part of the MetaBreast project by the Digital Surgery Lab.
"We are in the process of registering the company, which will be called Heka Vision," says Gouveia. "We have a five-year roadmap with investment needs of 13 million euros to carry out all the necessary clinical trials and put the device on the market", he adds. He also points out that, although the device's applications under development are still exclusively in the area of breast cancer surgery, its scope of use could obviously expand.
Why the name Heka Vision? It was Tiago Marques, a neuroscientist and computer vision specialist who is also affiliated to the Digital Surgery Lab and the MetaBreast project, who suggested it. Here, a parenthesis: in addition to the three senior researchers mentioned here, the Digital Surgery Lab has three PhD students, two software developers from an external company, working exclusively within our organization – and will soon include several master and undergraduate students. "There are already more than ten of us," says Gouveia.
Returning to the name of the future company, "Heka was the Egyptian God of medicine and magic," explains Marques. "So his name makes sense because we're developing medical devices. But we also like the fact that he's the God of magic, because we're developing a technology that basically works almost like magic, doesn't it?" Marques was inspired by science fiction and futurist Arthur C. Clarke, who said that a sufficiently developed technology seems like magic to people who don't know how it works. "The name Heka Vision translates well this idea of a medical vision that is at the same time magical", he points out.
