Table 2. Evaluation of classifiers as indicators of tumor infiltration.

Nuclear density, axonal density, and protein/lipid ratio were measured for each of the 1477 300 × 300–μm2 FOVs from 51 fresh tissue biopsies from 18 patients. A quasi-likelihood approach with a GAM was used to incorporate all of the attributes into a single classifier. Half of the FOVs (n = 738) were used to create the classifier, which was then tested on the other half of the data (n = 739). Given that glioma can be more difficult to distinguish from normal tissue than metastases and extra-axial tumors, a quasi-likelihood GAM was also used on a subset of tumors, excluding all nonglial tumors, to create the glioma-only classifier. To eliminate correlation between the testing set and training set, we used a leave-one-out cross-validation approach. The leave-one-out cross-validation was performed in a data set excluding nonglioma patients. CI, confidence interval; AUC, area under curve; n/a, not applicable.

Classification conditionAUCMean sensitivity (%)95% CIMean specificity (%)95% CI
GAM (all subjects)
  Normal versus abnormal0.99597.595.9–98.998.597.0–99.7
  Normal versus infiltrating0.98894.791.4–98.998.597.0–99.5
  Normal versus dense0.98998.095.6–10099.097.4–100
GAM (glioma only)
  Normal versus abnormal0.99497.095.0–98.798.797.2–99.5
  Normal versus infiltrating0.98894.991.3–98.198.597.1–99.5
  Normal versus dense0.99098.295.1–10099.098.2–100
Leave-one-out cross-validation
  Normal versus abnormal0.89387.3n/a87.5n/a
  Normal versus infiltrating0.91182.8n/a95n/a
  Normal versus dense0.90883.9n/a93.3n/a