Dekel Gelbman is the CEO of FDNA, a company that uses Artificial Intelligence to help patients with rare genetic diseases to get better disease management and treatments.
In this interview, Mr. Gelbman discusses FDNA and its commitment to improve outcomes for patients, as well as leveraging AI to help doctors handle huge amounts of unstructured data, achieve more accurate diagnosis and work in a more personal and human way.
What is your background and how was FDNA
FDNA’s story starts with Facebook almost 10 years ago. A different company founded by FDNA’s founders (face.com) had developed the best facial recognition technology in the world at that time and was sold to Facebook. After that success, we were looking for the next challenge, utilizing our expertise. We recognized that artificial intelligence can make a huge impact on healthcare. We started in 2011 and quickly realized that the main challenges were around obtaining data, integrating into doctors’ workflow and protecting patients’ privacy. Since our beginning, FDNA has established itself as a global digital health leader in the space of clinical genomics.
What is FDNA’s mission?
FDNA’s products and services seek to improve outcomes for patients, by supporting providers and facilitating more accurate and efficient genetic testing for clinical and molecular diagnostics.
Rare disease patients wait for 7.5 years on average to reach a diagnosis. This is mainly due to the complexity of diagnosing rare genetic conditions, lack of awareness and cost of diagnostic tests. This diagnostic odyssey impacts patients’ outcomes severely. If patients are diagnosed earlier, they could benefit from better disease management and treatments.
Being able to identify and describe accurate phenotypes can lead to higher diagnostic rates with greater efficiencies, which could lead to more efficient drug development and trial.
Walk us through FDNA’s products, and what problems they solve. Which product are you currently focusing the most on?
FDNA’s technology (next-generation phenotyping, or NGP) and products (Face2Gene Suite) are used globally by about 70% of the clinical geneticists across 2,000 different clinical sites in 130 countries. We do not charge clinicians for access to our technology or products. Our customers are primarily genetic testing laboratories who use our products and integrate our phenotype analysis into their variant analysis pipeline to achieve superior results when sequencing patients.
FDNA’s first NGP technology, DeepGestalt, is an AI-based module that analyzes facial photos of patients to highlight phenotypic similarity to known, but rare genetic disorders. This module was recently published in a peer-reviewed manuscript in Nature Medicine. It analyzes facial photos of patients to produce a list of syndromes that are found to have similar facial characteristics with accuracies of above 90%. DeepGestalt currently supports more than 300 specific genetic syndromes and syndrome groups, representing 45% of cases solved by whole exome sequencing.
We are now testing our technologies for clinical trial recruitment, as well as patient discovery projects. Being able to identify and describe accurate phenotypes can lead to higher diagnostic rates with greater efficiencies, which could lead to more efficient drug development and trial.
Which product do you think is making the biggest impact, and why?
A patient’s phenotype is a critical component in reaching a diagnosis. FDNA’s NGP technologies capture, structure and analyze human physiological data to produce actionable genomic insights. Our flagship product, Face2Gene, helps to reduce the cost of healthcare by increasing accuracy and cutting down the time it takes to reach a diagnosis for a genetic disease with significant efficiency.
Recently, Face2Gene also expanded to the variant analysis domain with Face2Gene LABS, launched in partnership with PerkinElmer Genomics. Face2Gene LABS uses the same AI to integrate the phenotypic score directly into a variant analysis platform (analyzing genome sequence data) and help filter and prioritize the list of variants found. One recent study suggested that this approach could increase the diagnostic yield by more than 400% compared to standard approaches.
We view all of our products as part of a continuum, representing the patients’ journey. The impact in incremental.
How is your product improving patient experience?
Patients living with a genetic disease often times go years without receiving a diagnosis, which means they are also not receiving the appropriate treatment or therapies that are available for some of these syndromes. By aiding in reaching a faster, more accurate diagnosis, patients can seek appropriate treatments and care that can improve their quality of life and extend their lives.
How do you measure the performance of your products?
There are many ways we measure performance. We measure the accuracy of the technology. We measure the usability of our products. We measure the number of publications using our technology. The most important one, in my opinion, is the number of lives we impact.
Join our weekly newsletter to receive:
- Latest articles & interviews
- AI events: updates, free passes and discount codes
- Opportunities to join AI Time Journal initiatives
Finding more patients for rare diseases will lead to more effective research and will bring better drugs to the market faster, which means better outcome for patients.
The most valuable impact of AI on the way patients will be cared for is by freeing up the doctors’ time and letting them go back to treating patients in a personal way. If you ask me, that is by definition personalized healthcare!
Tell us about the FDNA team.
We have a very diverse team comprised of AI researchers, engineers, data scientists, biologists, medical professionals, marketing and business professionals and they are located in various places around the world, mainly in Boston and in Israel. The one thing I can say about our team is that they are all driven by passion to make our world a better place and by a deep commitment to the rare disease community. I will say that an extension of our team is our Scientific Advisory Board – while not “officially” part of our team, they are as important to our efforts to develop our technology and products and introduce them in practice.
What are the primary issues in healthcare that AI can help to solve?
The ever-growing amount of data is a huge problem in healthcare. Combining that with the ever-increasing pressure and decreasing time doctors have for processing data negatively impacts patients’ outcomes. AI will be increasingly impactful for all practices that heavily rely on human interpretation of unstructured data, such as medical imaging. We are already seeing how AI is integrated and used by medical professionals to augment their analysis. Additionally, we see a growing use of AI in research and in drug discovery, which will support the advancement of precision medicine in the next few years.
AI will be increasingly impactful for all practices that heavily rely on human interpretation of unstructured data, such as medical imaging. We are already seeing how AI is integrated and used by medical professionals to augment their analysis.
How will AI affect the work of the doctors?
AI will become part of the doctors’ tool kit. Like how doctors use medical devices to obtain data that their human senses cannot acquire, so will they start relying on AI to perform analyses that the human brain cannot perform. Particularly, big data processing and pattern recognition. AI is simply superior than humans in those tasks. It is critical however to design and position AI systems with the value of augmenting clinicians’ capabilities by aiding in the aggregation of and sifting through mass amounts of data—never threatening to replace clinicians. When that becomes the norm, doctors will have more time for patient care and counseling.
How can doctors prepare for AI-powered healthcare?
That’s an excellent question. Doctors should start learning how AI-based solutions work and what data they are trained on. They need to respect AI’s capabilities but also think critically about AI-based analyses. Mostly, they need to embrace AI in the early stages, so they don’t find themselves left behind and practicing a different kind of medicine in a few years.
Do you think that AI will drastically change the way patients are cared for?
Yes, and for the better. Not only in simple measures of improved diagnostic yield and better trial design but in more cost-effective utilization of resources – those go without much saying. The most valuable impact of AI on the way patients will be cared for is by freeing up the doctors’ time and letting them go back to treating patients in a personal way. If you ask me, that is by definition personalized healthcare!
Which healthcare process has the biggest potential for improvement in the coming years and why?
Without a doubt, diagnosis. This is an area, where all the positive attributes of AI have immediate actionability. For example, in the realm of genetic testing, we already demonstrated a huge leap in diagnostic efficacy by integrating our next-generation phenotyping technologies into standard genetic sequencing interpretation, and almost tripling the diagnostic yield by doing so. Surely, other processes will be improved, even billing, but I personally believe the greatest impact will be in diagnostics.
I think that we are just starting to discover how deep and wide human genetics is related to every healthcare domain. Since genetics inherently involve analysis of huge amounts of data, I think that any AI technology in this field will have a huge impact on the entire healthcare domain.
What are the current technological limitations of AI which, once overcome, could bring massive improvements to the healthcare system?
I believe the largest challenge is the quality and availability of data. Garbage in; Garbage out – so really, an AI solution is only as good as the data it’s trained on. Sharing data, while protecting patient privacy, is one of the most important, interesting, ethical and legal challenges AI is facing when it comes to healthcare. This has a huge impact on equitability as well – AI should help us bridge the gaps in healthcare, rather than widen them. This is a topic that policymakers need to address and the sooner, the better.
Which AI-related technology trend do you think will have the biggest impact in healthcare in the coming years, and why?
While I’m biased, I believe that anything that relates to genetics will have a profound impact on healthcare. I think that we are just starting to discover how deep and wide human genetics is related to every healthcare domain. Since genetics inherently involve analysis of huge amounts of data, I think that any AI technology in this field will have a huge impact on the entire healthcare domain. Specifically, I think that the integration of sequencing with other omics data will yield interesting findings. We, at FDNA, are banking on phenomics, as one of the most important domains and anticipate that genomics will be a standardized integral component in all health-related decisions. Looking towards the future, I believe that every child will be sequenced and phenotyped, and that AI will significantly contribute to better outcomes for patients on a global scale.