Kidney stones are common, affecting more than 1 in 10 people, and these hard deposits made of minerals and salts can lead to severe pain, urinary tract infections, and loss of kidney function if they are left untreated.

Effective treatment and prevention rely on knowing what stones are made of. That is often determined through a manual process using FTIR. However, mistakes in FTIR interpretation and reporting can lead to inappropriate treatments and undermine efforts to prevent future kidney stones through medication, diet, and lifestyle changes.

The Mayo Clinic researchers conducted a study to determine whether their algorithms could find errors in kidney stone composition results that had been reviewed and reported by a clinical laboratory technologist.

Within a 12-month study period, the AI-trained algorithms reviewed manual analyses of 81,517 kidney stones. While the technologist's and AI's interpretations agreed 90% of the time, the researchers determined that manual entry errors and/or incorrect technologist interpretations before the use of AI occurred at a rate nearly eight times higher than after the start of AI use.

"By adopting AI, clinical laboratories can reduce errors in results sent to physicians," Day said in a statement. "This study is an example of how the fields of laboratory medicine, data science, information technology, and biostatistics can work together to improve patient care."

The developers will present the results of a study that describes their work on Tuesday in a Scientific Poster Session at AACC 2022.