Hyperspectral Imaging Supports Brain Neurosurgery

A hardware accelerator is used to speed up the hyperspectral brain cancer detection algorithm to achieve processing during the time of surgery, which was developed using a labeled dataset comprised of more than 300,000 spectral signatures. In a preliminary study, thematic maps of seven hyperspectral images of in vivo brain tissue captured and processed during neurosurgical operations demonstrated that the system is able to discriminate normal from tumor tissue in the brain within one minute.

To achieve this real-time discrimination, huge amounts of information captured by the sensors is processed using a K-Nearest Neighbors (KNN) filtering algorithm, which is optimized and parallelized by exploiting using graphical processing unit (GPU) technology for real-time processing during brain cancer surgical procedures. The parallel version of the KNN filtering algorithm can effectively handle the extremely high computational requirements needed to evaluate different classes simultaneously. The study describing the HIS development process was published in the July 2018 issue of Sensors.

“They, being the neurosurgeons, had a problem and we had a technology. But every patient’s tumor and brain produce a unique spectral fingerprint, and so the first algorithms to make usable images took a half hour; now the total time is around six seconds,” said study co-author Professor Gustavo Marrero Callicó, PhD, of ULPGC. “Now they are equipped to provide neurosurgeons with a tool to operate on the slimmest of margins in real-time. The next goal is refining the database to make it general enough to detect cancers in many situations.”

HSI can help acquire large numbers of spectral bands throughout the electromagnetic spectrum (both within and beyond the visual range) with a very fine spatial resolution. So fine, in fact, that for every image pixel a full spectrum of color can be detected. Using this information and complex classification algorithms, it is possible to determine which material or substance is located in each pixel.

Related Links:
University of Las Palmas de Gran Canaria
Imperial College London
Universidad Politécnica de Madrid