At this year's Upper Bound conference, AI experts, engineers, and entrepreneurs gathered to share the latest advancements and opportunities that artificial intelligence offers in their fields.

Many sessions explored how machine learning is transforming the way that healthcare works, offering new tools for medical professionals and better outcomes for their patients.

Generative AI in Healthcare

In her session, “Generative AI in Healthcare”, AI/ML engineer Jeevanshi Sharma laid out how generative AI could help tackle one of medicine’s most challenging problems: discovering effective new medications. 

 She explained that developing new drugs can be an expensive, time-consuming process. Pharmaceutical researchers often rely on a trial-and-error approach when designing new compounds, encountering many dead ends of ineffective or unsafe options before identifying promising candidates for new drugs. Generative AI could help reduce wasted time and money by better predicting how these compounds might work in the human body, allowing researchers to focus their resources on the most promising options.

“Machine learning could help scientists explore vast chemistry spaces in hours, instead of years,” she said.

This wouldn’t only make current drug discovery research more cost-effective, she notes, but it would also allow the development of new treatments for more complex and rare diseases, where it isn’t currently economically feasible with current technology.

While generative AI offers immense potential in drug discovery, Sharma cautioned about inherent risks. Current models can sometimes "hallucinate," producing false predictions. While problematic in many fields, this could be dangerous in medical settings.

To ensure safe use of generative AI in healthcare, Jeevanshi proposed several solutions. These include explainable AI systems that cite their sources and human-in-the-loop approaches where human experts validate AI models. She emphasized four crucial pillars for safe generative AI use in drug discovery: Transparency, Validation, Good Governance Practices, and Training.

Cutting Edge AI Advances in Healthcare and the Future of AI In Health

“We still haven’t scratched the surface of what’s possible, or what’s plausible,” said Dr. Zaahir Moloo during his session, “Cutting Edge AI Advances in Healthcare and the Future of AI in Health”, on the ways that AI is changing the way doctors track their patients’ medical data. 

Moloo, in addition to being a family physician in Edmonton, is the co-founder of Scribeberry, an AI-powered tool that automatically captures information when doctors interact with their patients. 

Even in its early stages, machine learning in medicine is already making a significant impact on critical problems facing healthcare professionals. A major hurdle is the lack of data interoperability in healthcare. Different clinics and practitioners use varied software, making seamless data transfer between Electronic Medical Records (EMR) systems difficult and often leading to errors or missing information.

“In healthcare right now, we are kind of siloed off. We have all kinds of different tools.”

AI-enabled tools can break down these data silos by facilitating information sharing between EMR systems while maintaining patient privacy. Moloo showcased Scribeberry, which not only streamlines data transfer but also automates tasks like new patient intake and even assists patients with applying for disability support and medical tax grants.

These are all steps towards a greater goal. Moloo envisions a sort of “healthcare OS,” that works much like the internet does, but for medical professionals. An AI-enabled system that would allow effortless communication of a patient’s information, reducing errors and allowing resources to be focused on other areas.

At the end of the sessions, Moloo highlighted some of the near-future machine learning applications that he sees on the horizon, including AI tools to identify signs of diabetic retinopathy during eye examinations, robotic surgery, and better monitoring of patient health in pediatric emergency medicine.

Application of AI In Dentistry

In his session, “Applications of AI in Dentistry”, Hollis Lai —  professor and Senior Director of Innovation and Quality Improvement at the University of Alberta’s School of Dentistry —  shared insights into the ways that dentistry differs from other medical practices and the unique opportunities that machine learning approaches would offer in the field.

Data collection in dentistry differs significantly from other medical professions. It's often held by individual dental clinics, each with its own methods. This means dental data is highly detailed at an individual level but challenging to aggregate. Machine learning projects utilizing dental data must account for these conditions.

Lai then highlighted two specific projects that showed the potential of machine learning in dentistry.

The first involved using deep learning techniques to analyze and classify images of patients’ midpalatal suture, which is a cranial joint at the top of a person’s mouth. This joint is very important for the growth of a person’s upper jaw and facial development, and knowing what state the joint is in guides dental treatment plans. In this study, a machine learning model was trained to automatically determine the maturity of a midpalatal suture with a high degree of accuracy, giving vital information that could lead to more effective treatments for the patient.

A second project involved using machine learning to predict a child’s dental anxiety. Dental visits can be stressful for patients of all ages, and Lai explained that dental offices often use surveys to determine how much anxiety a patient has. While this approach can work with adults, it isn’t an effective way for children to communicate their fears.

Instead, this project asked the kids to draw pictures while in the waiting room of the dentist’s office. These images were then analyzed by a model that ML techniques, including facial expression classifiers and human figure analysis, to provide insight into the anxiety that a young patient might be feeling about their appointment. Lai said that recent advancements in machine learning have made his kind of work possible, given the data challenges in dentistry.

“It shows we can take the data from a very small sample and do a lot with it.”

(Photo Credit: Ampersand Grey)