Having developed an algorithm that takes five biomarkers into account, the team hopes to detect tumour tissue with significantly higher accuracy than the current state-of-the-art: qualitative visual assessments of fluorescence contrast produced by endogenous protoporphyrin IX (PpIX) or even quantitative assessments of PpIX concentrations alone.

"We are, to the best of our knowledge, the first group to quantify and incorporate quantitative biomarkers into a biologically relevant approach for guidance during resection of gliomas," researcher Pablo Valdés from Dartmouth College's Thayer School of Engineering, told medicalphysicsweb. "This approach even extends to detecting low-grade gliomas in vivo with a high diagnostic accuracy."

From fluorescence to optical guidance

Fluorescence-guided resection involves administering the patient with a substance called 5-aminolevulinic acid (ALA). ALA produces PpIX in the body, which preferentially accumulates in cancerous cells and emits a characteristic red fluorescence when excited with blue-violet light. This combination of factors makes it ideal for guiding the resection of gliomas, which are notoriously diffuse in nature.

Previous work in the field of fluorescence-guided neurosurgery has relied simply on qualitative assessments of the levels of visible PpIX fluorescence, as seen by the naked eye, to help guide surgery.

"We hypothesized that if we quantified additional biologically relevant biomarkers which are indicative of pathophysiological changes in tissue, in addition to PpIX, we would further improve tumour detection accuracy," commented Valdés. "Fluorescence guidance then becomes true optical guidance through the use of quantitative fluorescence and other optically derived biomarkers."

Optically guided resection

The team used a specially designed probe containing four optical fibres to acquire all of the necessary information to feed into its algorithm. Two of the fibres transmit white light, one is used to transmit blue-violet light and the final fibre collects the emitted signal for analysis.

First, the probe is placed in direct contact with the tissue that the researchers wish to analyse. The site is then sequentially interrogated with white light to collect two reflectance spectra and then blue-violet light to collect one fluorescence spectrum.

"We then use the white light data in a light transport modelling algorithm that gives us a quantitative measurement of haemoglobin concentration, oxygen saturation and scattering parameters," explained Valdés. "This information also gives us the ability to correct for the distorting effects of tissue optical properties on the fluorescence emissions to measure the absolute concentration of PpIX and its associated photoproducts."

The estimates of each of these five biologically relevant markers are the input variables required by the group's diagnostic algorithm, which determines whether the tissue at the site in question is normal or abnormal.

"We used a multivariate classification algorithm that finds the optimal separation between two classes – normal and abnormal tissue," explained Valdés. "Each data point is described by the five biomarkers, which then undergo a mapping transformation from the 'raw input space' into a 'feature space' where the classification and separation of each tissue into their respective class is done."

Results and future work

The Dartmouth/University of Toronto group tested its algorithm on 10 patients undergoing fluorescence-guided resection. A total of 88 sites were interrogated with the probe, each in triplicate, resulting in 264 spectra acquired from 26 control and 62 tumour locations.

The researchers report that: "tumour tissue delineation achieved accuracies of up to 94% [specificity=94%, sensitivity=94%] across a range of glioma histologies, beyond current optical approaches, including state-of-the-art fluorescence image guidance."

"We wish to continue to test this multiple biomarker approach by evaluating it on a larger cohort," concluded Valdés. "Our group, led by Keith Paulsen, Professor of Engineering, and David W Roberts, Professor of Surgery (Neurosurgery) at Dartmouth, is currently working on extending this quantitative approach to a wide-field imaging modality, as opposed to a single spectroscopic point approach."