Quick contact info

Calle Pollensa, 6 2ª Planta, Edificio ECU, 2, 28290 Las Rozas de Madrid, Madrid

+ (34) 916 49 22 92 icon_widget_image info@medlabmg.com

Marcos Rubio: “I’m motivated to help with my developments”

Marcos Rubio, computer vision engineer at MMG.

Marcos Rubio is part of the Image Processing team of the Artificial Intelligence and Research Department of MMG. Together with Clara Soler, he is developing a system of recognition of dental pieces and pathologies based on algorithms. In addition, he dedicates part of his spare time to a tele-assistance project with which he has won the Cerner #HCESmartApp award and the General COM SALUD / AIES of the Health Hackathon held recently in Madrid.

The Health Hackathon is the largest health programming event in Spanish. Along with Maria Gonzalez, Carlos Santiago and Ana Maria Sanchez, Marcos Rubio presented with the Live Health project. “Maria and I did the career together and were together in the residence,” says the computer vision engineer, who studied Biomedical Engineering at Carlos III University. Last July, his classmate set up a start-up to offer tele-assistance to elderly people.

Increase adherence

Gradually, the idea was defined, until it became a system of automatic calls to increase adherence to treatment. In order to do this, the system reminds the person to take medication, as well as asking how he or she is doing. In this way, in addition, a follow-up is available for family members for their tranquility. Now, Marcos Rubio and his team have adapted the system to the hospital environment so that it can be incorporated into the computer system of Cerner, one of the sponsors of the hackathon.

Marcos Rubio and his team during the Health Hachathon.

Marcos Rubio and his team during the Health Hackathon.

“The idea is the same because the problem is the same: adherence, which leads to emergency admissions or interventions that should not occur,” explains the engineer, according to whom his system saves costs through adherence. At the same time, it responds to a social need. “I have always been very interested in the social issue, that the developments you make reach people and help. Here at MMG, in the end, the developments we make are going to reach people. To me that’s what motivates me in the end, to help with my developments”.

“I believe that artificial intelligence is a good tool to try to reach people”, says Marcos Rubio, who recognizes that this is the force that moves him.

He is also motivated by his passion for numbers. In fact, he worked in a cancer research laboratory for 3 months, but ended up opting for the most technical facet of his training. “In the end I have always been orienting myself to the part of the algorithm because I like mathematics a lot”, says the engineer, who after the university degree did a Master of Multimedia and Communications while working for PlenOptika, a spin-off of the Massachusetts Institute of Technology.

Prescription eyeglasses

There, he developed the algorithm that made it possible to prescribe glasses in 10 seconds using a mobile autorefractometer. “A device like this costs between 15 and 20,000 euros and cannot be moved”, explains Rubio. In countries like India there are a lot of people who have needs, but they don’t have access to the diagnosis. With this mobile device, the aim was to facilitate access to prescriptions through the use of artificial intelligence.

“The device had cameras that measured the eye with a laser. It hit the retina, reflected, and that reflection was recorded with a camera -he says-. What we did was process the images in real time to get the prescription out of the glasses. There were no computers or clouds, so the code had to be highly optimized and work in real time. My responsibility was to make sure that all that algorithm worked well”.

“When we had done everything, we were selling and it started to be a more boring job, I decided to change so as not to stop doing things”, says the engineer, who then arrived at MMG, where he works with CBCT images. “The problem with these images is that they are very heavy and the doctor has to go tooth by tooth to examine them. The average evaluation time is 10-15 minutes. The goal was to make an algorithm that would do to you what a doctor does to you in 10 seconds. That way, the doctor checks only the alerts”, he explains.

Dental image

At present, the system is able to identify teeth with 95% success. In addition, it is capable of distinguishing 14 pathologies. “There are a lot of companies dedicated to image recognition. Now what is most talked about is the subject of radiography. Even algorithms get better than a set of doctors. It’s something that’s moving and that’s going to be there sooner rather than later in daily practice”, says Marcos Rubio.

“If you are given an image where a doctor is able to see something, the algorithm will also be able to see it if you train it well. If they give you a black image, you’re not going to get a cancer out of there”, says the engineer, who makes it clear that artificial intelligence is not magic.

In addition, he emphasizes the need to democratize the data (“of course, all anonymized”) so that the algorithms of medical imaging reach the consultations. “It is already done in the United States in a lot of clinics. In Israel, which is the strongest country in this, it is part of the routine. In Spain it’s being more complicated, but I wouldn’t be surprised if in the next 2-3 years it’s part of the hospital routine. That a machine makes a symtom checker of the image and tells you: check this that you’re missing something,” predicts the expert.