Researchers have shown that hyperspectral imaging and artificial intelligence could predict the presence of cancer cells with 80 to 90% accuracy in 293 tissue specimens from 102 head and neck cancer surgery patients. The smart surgical microscope could help surgeons better detect cancer during surgery, reduce operating time, lower medical costs, and save lives. Hyperspectral imaging is noninvasive, portable, and does not require radiation or a contrast agent.
Currently, pathologists analyze tissue samples from a patient who is undergoing cancer surgery and still under anesthesia in a process called intraoperative frozen section analysis. Several resections may be needed during a procedure as surgeons try to reach tissue with clear, or noncancerous, margins. In some cases, cancer cells cannot be sampled or detected during surgery, resulting in additional surgery.
Hyperspectral imaging — originally used in satellite imagery, orbiting telescopes, and other applications — goes beyond what the human eye can see as cells are examined under ultraviolet and near-infrared lights at micrometer resolution. By analyzing how cells reflect and absorb light across the electromagnetic spectrum, experts can get a spectral image of cells that is as unique as a fingerprint.
The researchers look to “train” the microscope to recognize cancer with the help of images of cancerous and noncancerous cells in an extensive database. Once trained, the smart device can predict whether a new sample is a cancerous tissue or not.
The technology should be able to provide nearly instant results, which could significantly reduce surgery time and cost. Each frozen resection evaluation under current procedures can take 30 to 45 minutes. In many cases, multiple evaluations are needed, which further prolongs the surgical time and increases the costs.