Scientific start-ups: software medtechs – FrenchWeb.fr

by bold-lichterman

For the previous articles, I had to go about it more times and correct some inaccuracies, as with what concerned Abbelight and I’nside. This pisses me off because I try to be as specific as possible in these articles. And that can rightly annoy the companies concerned! The amount of information collected and compiled, plus typing errors, and sometimes technical confusions generate this kind of unforgivable error. But you are not on a fake news: I obviously correct these unintentional errors as soon as I know about them!

In this new part on health, I will deal with some scientific software start-ups in the field of bioinformatics, in particular those in the genomics sector and which complement the medtechs in the field explored in a previous article.

In this series of articles, I use the same approach as in the CES Report: to carry out a more or less broad sampling of the market offers to deduce from them in “bottom-up” mode some trends and “patterns” in the markets. factors of success or failure of start-ups.

Medtech software

The software medtechs I am talking about here have a certain scientific heritage. They often analyze the results of medical devices, EKG systems such as medical imaging or from genomic analyzes. Machine learning is often present, even if it is less marketed than in digital start-ups.

ezratty

Let’s start with Cardiologs, a start-up specializing in the automatic analysis of electrocardiograms. I had the opportunity to meet Yann Fleureau, his co-founder on several occasions, and in particular during a visit to his start-up hosted in the premises of the Faculty of Medicine in Paris in November 2016. He and his co-founders are from the Ecole Polytechnique. Yann did a master’s degree in entrepreneurship at l’X. The company embarked on the analysis of ECGs by wanting to advance the state of the art. Until now, this was based on the analysis of the signal based on Fourier transforms, expert systems, peak detection and measurement of the durations of each phase. Some serious pathologies can be detected with the naked eye but others are more difficult to detect, especially by general practitioners.

ezratty

Cardiologs offers a cloud software solution that analyzes ECG data performed according to the rules of the art with several electrode patches (4 on the limbs, 6 on the thorax) in a doctor’s office, by nurses or doctors. The results are provided on a web interface. It is based on machine learning methods using convolutional neural networks with supervised learning (CNN). On the cloud side, they use the resources of Google Tensorflow as a number of AI start-ups. This allows a response to be provided in near real time. They trained their system with ECG bases including a 100,000 ECG base from Minnesota recovered in 2015. You have to pay for it, but it’s not too expensive.

They actually built two solutions, one for capturing ECGs at rest with the usual electrodes, and the other in an outpatient setting, using data from simpler sensors, such as smartwatches or even connected clothes. Their system is able to predict a hundred disorders on 12 channels (ECG at rest) and about 15 on 1 to 2 channels (ambulatory ECG). Knowing that the analysis of the coronary heart – therefore the risk of infarction – requires several electrodes, and if possible a stress ECG (where the patient pedals on a bicycle for 5 minutes). The system detects in particular the atrial fibrillation, which is correlated with the onset of stroke – cerebrovascular accidents – due to poor circulation of the blood, the risk of which increases with age and which is easier to treat when it is early.

To date, the start-up comprises a dozen people, with a balance of skills between mathematics, machine learning and web development. They also have the cardiologist Pierre Taboulet on their team, who plays the role of CMO – chief marketing officer – and is also at the origin of the site for learning to read ECGs e-cardiogram.com. They market their software mainly to reading centers which analyze them. Their solution makes it possible in particular to expand the ECG market by making it accessible to more practitioners, in parallel with the profession of traditional cardiologists. This increases the ability to make early diagnosis of cardiac pathologies. To date, the company has been funded to the tune of 1.2 million euros, including 200,000 euros in a grant from the Concours Mondial de l’Innovation, an honorary loan of 90,000 euros from Scientipôle and the rest from ‘a dozen business angels. It’s still light and they’re going to have to put on overdrive to develop.

AEDMap is a Parisian start-up crossed at the Web Summit in November 2016. Its Visio Plus solution makes it possible to map connected cardiac defibrillators and to supervise their maintenance in a predictive manner. This is perhaps where we can find a little specific technical material in their offer, the rest of the solution falling under a traditional management tool, for fleet management, its mapping, device traceability. and the management of their consumables, the history of their use and maintenance operations. They already have a few references with major clients such as Enedis and Conforama. These defibrillators are found in public places, transport as well as in businesses.

DreamQuark is a French start-up that creates healthcare data analysis solutions using deep-learning. In particular, she created the DreamUp Vision spin-off which automatically analyzes the fundus of the eye. This serves to identify the emergence of retinopathies in diabetics as early as possible. The average increase in the level of glucose in the blood has the effect of damaging the small blood vessels including those that supply the retina (in addition to those of the kidney, heart, and extremities of the limbs). The method used is a mixture of image processing and deep learning. It facilitates and speeds up the detection work of ophthalmologists. And since there are too few of them and this examination is recurrent, at least once a year and per patient, it is of great benefit to the community. Unlike other AI applications, this one does not risk destroying jobs! Machine learning-based techniques for medical imaging analysis will radically transform healthcare professions in the years to come!

ezratty

Portable Genomics is a start-up created by French people based in San Diego, USA since 2011. It has created a mobile software solution for collecting and visualizing a person’s health data. It ensures their online storage, under the control of the user. The solution collects both genomic data from sequencing (complete genome) or genotyping (analysis of typical variations in genes) as well as those from health in general: history of pathologies, lifestyle and data from connected objects. This makes it possible to constitute a 360 ° view of the patient, essential both for practitioners and for creating health databases that can be used by research and pharmacy companies. This also makes it possible to identify the level of risk of various pathologies. The company is positioned as a platform for the collection, sharing and monetization of personal health data, based on a revenue sharing model with users.

ezratty

The solution is in beta-test with groups of patients (in cancerology or rare genetic diseases) in conjunction with the pharmaceutical industry to accelerate research into new therapies. The company, like many genomics startups in Silicon Valley, 23andme first, assumes that health data is much more valuable than the machines that capture it, such as sequencers that are falling in price. keep on going. But for this data sharing to become widespread, it must be done through a relationship of trust with patients, both in terms of respect for privacy and the monetary value of this data. A very fine ambition!

Traaser is one of the few French companies in software associated with genomics. She created Diagen, an expert system for analyzing genomes to aid diagnosis and the choice of targeted therapies, in particular in the context of orphan diseases or of multifactorial origins. Created in 2016, it is based at the Génopole d’Evry. The company has many competitors in the United States, in particular applications in the same field carried out with IBM Watson by various American research institutes. The solution will be marketed in the cloud for physicians. It promotes technologies from the CEA Institute of Genomics. It will now have to find the means to finalize its offer and finance its commercial deployment. The marketing cycle should be a little shorter than for biotechs!

GenoSplice is a company that operates in the same field as Traaser but is much more specialized. Also based at the Génopole d’Evry, the company provides bioinformatics services for the analysis of DNA splicing. Késako? Splicing is the step of eliminating the non-coding sequences of genes (introns) in eukaryotic cells after their transcription into pre-messenger RNA, which gives a mature messenger RNA taking up the coding sequences of the exos and which can feed the ribosomes that will use the RNA code to generate polypeptides and proteins (polypeptides with a large number of amino acids). Are you still following? The diagram below provides a better understanding of the principle of splicing. It turns out that the same gene can be spliced ​​in several different ways depending on the influence of various regulatory factors that can reverse the position of exons or remove them. The company allows the study of these variations, which are associated with various pathologies, in particular in the emergence of cancers.

ezratty

osimis.io offers the open source medical imaging software of Belgian origin Orthanc, created by Sébastien Jodogne and Alain Mazy. The approach is original and unusual in this sector. The idea of ​​the creators is to provide a completely open source and free DICOM software solution (Digital Imaging and COmmunication in Medicine), allowing to build a server infrastructure for sharing medical and health imaging data, usable in particular by hospitals. It is operable with many proprietary medical imaging systems

ezratty

The solution is the result of research work at the CHU de Liège. Orthanc has won many awards in the open source world, including the Free Software Foundation, which is a bit like the Vatican of free software. Osimis.io is a company which, in line with open source business models, provides a range of services around Orthanc, for training, installation and maintenance. This brings to mind, in another area, the open source software VLC (media player) created by the students of the Ecole Centrale, which is now managed by an association (VideoLAN). Its president, Jean-Baptiste Kempf, is also the creator of Videolabs which develops and deploys solutions around VLC, in general, in service delivery mode. For Videolabs as for Osimis, direct wealth creation is low, but on the other hand, the impact for the entire ecosystem is significant. It’s a lifestyle choice!

This was just a sample. There are plenty of other companies specializing in medical IT. Some are not start-ups in the original sense of the term, but rather service providers.

Olivier-EzrattyOlivier Ezratty is a consultant in new technologies and author ofFree opinions, a blog on digital media (digital TV, digital cinema, digital photography) and on entrepreneurship (innovation, marketing, public policies, etc.). Olivier is an expert for FrenchWeb.

Read also: Scientific start-ups: imaging and operative medtechs