Artificial intelligence may play role in refractive surgery diagnostics
Click Here to Manage Email Alerts
NEW ORLEANS — Artificial intelligence may be used as a refractive surgery diagnostic tool, according to a speaker at Refractive Surgery Subspecialty Day at the American Academy of Ophthalmology meeting.
“A robust [machine learning] process generally includes building new variables, which means feature engineering, choosing the most appropriate algorithm, optimizing its parameters, selecting the most predictive features, and understanding the interconnection and patterns between both existing and created selective variables,” Marcony R. Santhiago, MD, PhD, said. “Thus, a better identification of patients at higher risk becomes possible regardless of the cutoff point associated with each one of the features.”
Santhiago described the use of AI in medicine as “human-computer systems,” noting the role people play in useful AI includes creating and maintaining the software, selecting which application to use and fixing problems.
To create useful AI, he said researchers should ask two key questions.
“What tests should the computers do, and what tests should people do? How can these human-computer systems improve over time?” Santhiago said.
Santhiago used an AI tool in a previous study to identify early forms of keratoconus. He said that AI can be used as a diagnostic aid for specific diseases by recognizing topographic patterns, finding interactions between topographic and tomographic indexes, and detecting early signs of disease evolution.
Using the AI model, Santhiago and colleagues were able to obtain 100% sensitivity for identifying early keratoconus.
“In our decision-making flowchart, AI can contribute to better identifying early forms of keratoconus when we need additional data and to better understanding the surgical impact on healthy corneas,” he said.