The majority of work6,12,13,16
using fluorescence spectroscopy as a tissue characterization tool (for example, in brain tumor studies) has been based on measurements of the raw emission spectrum without correcting for variations in tissue optical properties or the presence of other fluorophores. To assess the added value of our quantitative measurements, raw spectroscopic variables were also considered and their relative diagnostic performances were evaluated. Specifically, the variables A615
, and P710
were computed from the raw fluorescence emission spectrum.
Measurements were obtained under an institutional review board–approved protocol with informed consent during the open cranial surgeries in 14 patients with a range of tumor histologies: LGGs (in 2 patients), HGGs (in 3), meningiomas (in 6), and carcinoma of the lung metastases (in 3). In vivo spectra and corresponding tissue biopsies were recorded at distinct stages of the surgical procedure and were identified as corresponding to either the center or the edge of tumor in an effort to maximize the sampling extent of each lesion. Tissue samples from each biopsy site were also assessed histologically using H & E staining (). We found a statistically significant increase (p < 0.05) in CPpIX across all of the tumor categories when compared with the normal controls. presents CPpIX on a logarithmic scale for each tissue category, since the normal control values of CPpIX can be up to 4 orders of magnitude smaller than in tumor. The other diagnostic variables—A615, A660, P635, and P710—did not show similar potential for discriminating between normal and abnormal intracranial tissue.
Fig. 1 In vivo spectroscopic measurements of ALA-induced PpIX fluorescence during intracranial tumor resection surgeries. Left: Intraoperative fluorescence images of the resection cavity visible to the surgeon through the operating microscope under blue light (more ...)
Fig. 2 Concentration of ALA-induced PpIX, CPpIX. Circles represent the CPpIX value calculated in vivo using the light-transport model for each location where measurements were collected, with filled circles representing interrogated sites with visible red fluorescence, (more ...)
We used ROC analysis15
to further assess the diagnostic performance of the fluorescence variables listed in . We found that CPpIX
stood out as the most accurate diagnostic variable based on an AUC metric. In fact, CPpIX
discriminated abnormal from normal tissue with a mean AUC of 0.95 ± 0.02 compared with mean AUCs of 0.54 ± 0.06, 0.54 ± 0.06, 0.60 ± 0.06, and 0.57 ± 0.06 (± SE) for A615
, respectively. As summarized in , ROC analysis of CPpIX
as a diagnostic biomarker resulted in classification efficiencies of 87% for all tumors, 76% for LGGs, 93% for HGGs, 97% for meningiomas, and 95% for the metastases group ().
Receiver operating characteristic curve analysis of each diagnostic variable in the 5 categories of pathogenic tissues*
Summary of ROC analysis of CPpIX as a diagnostic variable*
Fig. 3 The ROC curve analysis of intraoperative detection of ALA-induced PpIX. A: Curve for all tumors using visible in vivo fluorescence as a diagnostic variable (AUC = 0.73 ± 0.03). B: Curve for all tumors using quantitative in vivo PpIX concentration, (more ...)
State-of-the-art clinical detection of PpIX during open cranial tumor resection is based on broad-beam blue light illumination and human visual perception and/or image capture (with a charge-coupled device) of the resulting fluorescence observed through the optics of the operating microscope. We have compared the sensitivity and specificity of this qualitative visual imaging approach with the quantitative fluorescence measurements presented here in the same cohort of patients (). Specimens were assigned a fluorescence score from 0 to 4 (0, no fluorescence; 1, minimal fluorescence; 2, moderate fluorescence; 3, high fluorescence; and 4, very high fluorescence) based on the impression of the surgeon (blinded to the quantitative measurement) of the visible fluorescence before the tissue was removed. The optimal classification efficiency was 66% (specificity = 100%, sensitivity = 47%, PPV = 100%, NPV = 51%, cutoff value: fluorescence score = 1, that is, minimal level of observed fluorescence) when using the surgeon's visual assessment, compared with a classification efficiency of 87% (specificity = 92%, sensitivity = 84%, PPV = 95%, NPV = 77%, cutoff value: CPpIX= 0.0074 μg/ml) when using the quantitative fluorescence measurements in the all tumors category. Furthermore, more than 81% (57 of 70) of the quantitative fluorescence measurements that were below the threshold of the surgeon's visual perception were classified correctly in an all-tumors analysis. shows ROC curves comparing the qualitative visual approach with the quantitative CPpIX data, which is significantly more accurate (quantitative approach: AUC = 0.95 ± 0.02, visible approach: AUC = 0.73 ± 0.03; p < 0.0001).