HEALTHCARE

When it comes to healthcare data, physicians and medical researchers face a common dilemma: analyzing large datasets can yield valuable insights, yet HIPAA regulations require an absolute dedication to preserving patient privacy. Galois used advanced cryptographic techniques, including multi-party computation (MPC), private set intersection (PSI), and differential privacy, to empower organizations to responsibly harness data insights garnered from aggregate statistics, while safeguarding the personal data of individuals, ensuring regulatory compliance and mitigating risks of privacy breaches.

Hospitals are complex technological ecosystems, stuffed full of a constantly changing network of interconnected medical devices, healthcare applications, and commodity enterprise IT. That complexity and interconnectivity risks patient safety by exposure to cybersecurity threats, limits access to care due to overloaded resources, and increases provider fatigue due to alarm showers. Galois has extensive experience developing tools for analyzing complex cyber-physical systems for government agencies like DARPA, NASA, DHS, and the American military. Many of these same technologies, used on incredibly sophisticated military aircraft and nuclear power plants, are just as effective in the healthcare domain – empowering healthcare professionals to understand and take control of their technological ecosystems, design safe and secure medical devices, and secure their hospitals from threats.

Human bodies are all unique ecosystems – no two are exactly the same. Variation in genetics, environment, lifestyle, diet, and more impact how individuals respond to treatments, medications, and interventions. While generalized solutions work well in most scenarios, more complex medical challenges often require, or could benefit from, treatments tailored to the distinct needs of the individual.
‍
Machine learning (ML) offers a potential solution – empowering doctors and researchers to uncover patterns in the vast oceans of data and constantly changing variables at work in a human body, and create application-specific solutions for health conditions that cannot be adequately managed with a one-size-fits all approach. 

Galois’s team of AI/ML researchers are actively exploring integrating high-assurance formal methods into machine learning-driven healthcare solutions. From improving automated insulin-dosers to developing personalized interventions for multiple sclerosis patients to using ML-systems as collaborative teammates in high-risk medical scenarios, such as providing real-time image analysis for robotic surgery.

Formal verification and mathematical modeling, combined with machine learning, also offer promising possibilities for replicating medical trials in a virtual setting. Combining these techniques and fields, we can create a “digital map” to enable reasoning about infinite patient scenarios (device efficacy for different individuals, edge cases, safety, etc.) before going to clinical trials. These virtual patient trials can be used to improve the safety of medical devices, catching problems quickly and cheaply prior to large-scale clinical trials, decreasing risk for manufacturers and healthcare providers, and improving outcomes for patients.