Velmio is a startup that I co-founded to build AI-powered tools for women’s health issues that have historically been underrepresented in medical research and product development. Women’s health encompasses female-specific conditions, as well as conditions that affect women differently or disproportionately. At Velmio, we built digital health apps integrating data from wearables and other sources with AI technologies to provide women with tailored analyses of their health.

Digital health applications for women's health

More than reproductive health

"Women's health" is typically associated with reproductive health conditions, such as menstrual disorders and pregnancy complications, as well as women's oncology (e.g., breast cancer, ovarian cancer). However, women's health encompasses a broader range of health challenges, including conditions that more commonly or more severely affect women (e.g., autoimmune diseases, migraines, osteoporosis and cardiovascular disease). Moreover, gender bias in care delivery can lead to underdiagnosis, undertreatment, or misdiagnosis of women's symptoms, especially in the areas of pain management and mental health.

The need for better women's health solutions

Women's health issues have been historically underrepresented in medical research and product development. For example, early research in cardiovascular disease largely focused on male subjects. Today, women are 50% more likely to be misdiagnosed following a heart attack and are more likely than men to die from a heart attack [1]. Women have been historically underrepresented in clinical trials, and conditions that more commonly affect women receive less funding for medical research - even for conditions with a significant burden of disease [2]. Consequently, women are twice as likely as men to experience adverse events from drugs [3].

Building a digital health platform for women's health

At Velmio our mission was to build digital health applications that would improve the patient experience and patient health outcomes for underrepresented groups. The first product that we developed aimed to reduce the risk of health complications during pregnancy.

The "data gap" in women's health refers to the lack of adequate and accurate data on various aspects of women's health, such as their specific needs, risks and outcomes [1,4]. We saw this problem as an opportunity to build a digital platform for collecting and analyzing high quality datasets for women's health, to help patients and their healthcare providers achieve personalised and evidence-based care.

To this end, we built a mobile health (mHealth) application to securely connect patient data from hundreds of sources (wearables, dietary logs, etc.), creating a "digital twin" of a patient's health and lifestyle. We then implemented machine learning algorithms to facilitate prevention, early detection and monitoring of health issues​.

Technological achievements in the Velmio pregnancy health app included:

• Connecting high quality data sources that had not previously been analysed together in a mobile application in the pregnancy health space, including digital biomarkers derived from consumer wearable devices to track the physiological changes that occur during pregnancy and environmental data to measure the impact of air pollution on pregnancy health

• Developing advanced natural language processing (NLP) technology to replace generic health information with insightful and personalised advice. For example, during the COVID-19 pandemic we analysed social media content to identify sources of stress at each stage of pregnancy, so that our app could better support the mental health of pregnant women

• Providing digital health tools to help patients manage pregnancy complications, such as gestational diabetes, pre-eclampsia, and severe morning sickness. Velmio was the first pregnancy app to offer a dedicated set of tools for patients with gestational diabetes, the most common pregnancy complication

• Developing AI/ML tools to improve the user experience for health tracking. This included our own on-device image recognition and semantic search models for nutrition tracking, as well as a machine learning tool that analysed ingredient composition and medical guidelines to indicate foods that are safe to eat during pregnancy

The Velmio pregnancy health app was utilised by over 30,000 people across 90 countries. We later extended our product offering to a second app for pregnancy care providers (including doulas, midwives), to create a seamless interaction between healthcare providers and our patient-centric solution.

Our work in the digital health space garnered global attention, being featured by the likes of Sifted/Financial Times, CNN, and EU-Startups. We also repurposed our technology to build one of the first COVID-19 health apps in the world, which was a winning entry of Estonia's national coronavirus hackathon (Forbes article) and featured by the World Health Organisation (WHO).

A more general problem

The women's health challenges I was working on at Velmio relate to a broader issue in health and medicine - solutions designed for a general patient population often fail to meet the needs of specific patient groups. This realization aligns with the rising interest in personalized health and precision medicine approaches across the healthcare and pharmaceutical industries. These approaches aim to tailor medical treatments, practices, and decisions to individual patients based on their genetic profile, lifestyle, environment, and other personal factors.

As I reflected on the technology we were building at Velmio, I became motivated to dive deeper into the development of AI-based precision medicine technologies. This led me to pursue these ideas as PhD research.

With the emergence of large-scale health and biomedical datasets (e.g. population-wide electronic health record systems and genetics biobanks), opportunities for AI include:

• Integrating vast amounts of high-quality data for diverse populations from various data sources: genomic information, environmental factors, real-time health data from wearables and IoT devices, electronic health records, etc.

• Moving away from a one-size-fits-all approach to greater personalisation in medicine and healthcare delivery, e.g., machine learning tools to aid patients and healthcare providers with tailored diagnosis and treatment planning

• Advancing the fundamental understanding of biological mechanisms of disease with deep learning and other computational methods for omics data (genomics, proteomics, metabolomics, etc.)

At scale, AI/ML technologies for precision medicine can potentially benefit many diseases and patient groups. However, despite the promise, more work is needed to ensure that such AI/ML systems are fit-for-purpose. This is why my research has explored how to make AI/ML work better for individual patients, motivated by the types of health problems we were trying to solve at Velmio.

References

[1] Caroline Criado Pérez, Invisible Women: Data Bias in a World Designed for Men

[2] Kerri Smith, Women’s health research lacks funding – these charts show how

[3] Irving Zucker & Brian J. Prendergast, Sex differences in pharmacokinetics predict adverse drug reactions in women

[4] World Health Organisation, Closing data gaps in gender, March 23 2023