Boston: Frequently, the surgical procedure is not predetermined when a patient undergoes surgery to remove a tumor or treat a disease. The surgeon must know the margins of a tumor, its stage, and whether a lesion is malignant or benign to determine how much tissue should be removed. These decisions frequently depend on the collection, analysis, and diagnosis of disease while the patient is on the operating table. Speed and precision are crucial when surgeons send samples to a pathologist for analysis.
The findings are published in Nature Biomedical Engineering.
The current gold standard method for tissue examination is often time-consuming, and a faster method that includes tissue freezing can lead to artifacts that can make diagnosis difficult. A more effective method was developed in a recent study by researchers at the Mahmood Lab at Brigham and Women’s Hospital, a founding member of the Mass General Brigham health care system, and associates at Bogazici University. The technique uses artificial intelligence to translate between frozen sections and the standard approach, improving image quality to increase the accuracy of rapid diagnoses.
“We are using the power of artificial intelligence to address a longstanding problem at the intersection of surgery and pathology,” said corresponding author Faisal Mahmood, PhD, of BWH’s Division of Computational Pathology. “Making a rapid diagnosis from frozen tissue samples is challenging and requires specialized training, but this type of diagnosis is a critical step in the care of patients during surgery.”
To make the final diagnosis, pathologists use formalin-fixed, paraffin-embedded (FFPE) tissue samples; this method preserves tissue in a way that produces high-quality images, but the process is laborious and typically takes 12 to 48 hours. For rapid diagnosis, pathologists use an approach known as cryosection, which involves rapidly freezing tissue, cutting sections, and viewing these thin slices under a microscope. Cryosection takes minutes instead of hours, but it can distort cellular detail and compromise or tear delicate tissue.
Mahmood and co-authors developed a deep learning model that can be used to translate between frozen sections and more commonly used FFPE tissue. In their paper, the team showed that the method could be used to subtype different types of cancer, including glioma and non-small cell lung cancer. The team validated their findings by recruiting pathologists for a study of readers in which they were asked to make a diagnosis from AI images and traditional cryosection images. The AI method not only improved image quality, but also improved diagnostic accuracy among experts. The algorithm was also tested on independently collected data from Turkey.
The authors note that prospective clinical studies should be conducted in the future to validate the AI method and determine if it can contribute to diagnostic accuracy and surgical decision-making in real hospital settings.
“Our work shows that AI has the potential to make time-sensitive and critical diagnosis easier and more accessible for pathologists,” Mahmood said. “And it could potentially apply to any kind of cancer surgery. It opens up a lot of possibilities to improve diagnosis and patient care.”
