- Doctors have developed a deep learning algorithm that can cut the time needed to analyze and classify a tissue sample during surgery from 30 to 40 minutes to just 3 or 4.
- AI is sparking advancements in all areas of the healthcare, and as the technology advances, so will its ability to help doctors keep us healthy.
A Better Way to Diagnose
A study from the University of Michigan Medical School and Harvard University has concluded that artificial intelligence (AI) can help doctors diagnose brain tumors more quickly and more accurately. The team of physicians and scientists used deep learning to analyze 100 different brain tissue samples and accurately classify most of them into several broad categories.
After the doctors collected tissue using a method they dubbed stimulated Raman histology (SRH), the deep learning algorithm analyzed it. Normally during surgery, surgeons must pause the surgery for 30 to 40 minutes to allow time to process, freeze, and stain the brain tissue in a lab. Using the SRH technique and deep learning algorithm, the surgeons were able to diagnose tumors without leaving the operating room at all. Most importantly, they could cut the analysis time to just 3 to 4 minutes. This almost 90 percent decrease in diagnosing time can significantly reduce a patient’s risk for complications during surgery.
Currently, the deep learning algorithm can only identify tissues in four categories. However, Dr. Daniel Orrigner, first author of the study, aims to expand its capabilities to include eight different categories, which would cover almost all of the kinds of tumors that neurosurgeons encounter.
The Doctors of Tomorrow
Thus far, the technique has been tested on around 370 patients, with a major milestone set at 500. Right now, it has a 90 percent accuracy rate for tissue sample diagnoses. While this is less than the 90 to 95 percent accuracy rate of traditional methods, Orringer hopes he can increase the system’s accuracy over time as the deep learning algorithm will grow more accurate in its diagnoses as it has more date from which to learn.
While the current prototype is made for research purposes only, the team hopes to conduct a large-scale clinical trial to further test the program’s capabilities. They are particularly eager to bring the algorithm to small hospitals in remote parts of the country that have little or no access to neurologists. Although 1,400 U.S. hospitals perform brain tumor surgeries, only 800 board-certified neuropathologists are at work in the country. As this technology reaches more hospitals, the deep learning system can collect greater amounts of data and ensure more accurate diagnoses.
This isn’t the first incidence of AI in medicine, and definitely won’t be the last. In other instances, AI has been able to correctly diagnose patients when physicians couldn’t, identify cancer with the same accuracy as doctors, and even prevent blindness.
As neural networks continue to develop, their relevance in the medical field will only continue to expand, helping physicians keep us healthy.