Correlation between tumor oxygenation and 18F-fluoromisonidazole PET data simulated based on microvessel images

Assessing hypoxia with oxygen probes provides a sparse sampling of tumor volumes only, bearing a risk of missing hypoxic regions. Full coverage is achieved with positron emission tomography (PET) using the tracer (18)F-fluoromisonidazole (FMISO). In this study, the correlation between different FMISO PET imaging parameters and the median voxel PO2 (medianPO2) was analyzed. A measure of the median PO2 characterizes the microenvironment in consistency with probe measurements.

Tissue oxygenations and FMISO diffusion-retention dynamics were simulated. Transport of FMISO and O2 molecules into and out of tissue was modeled by vessel maps derived from histology of head-and-neck squamous cell cancer xenograft tumor lines. Parameter sets were evaluated for 300 distinct 2 × 2 mm(2) vessel configurations, including medianPO2 and two FMISO PET parameters: FH denotes the sub-regional signal four hours post injection (pi) and FH/P denotes the ratio between FH and the time-averaged signal 0-15 min pi. Correlations between O2 and FMISO parameters were evaluated. A receiver operating characteristics (ROC) analysis was performed, regarding the accuracy of FH and FH/P in identifying voxels with medianPO2 < 2.5 mmHg.

In hypoxic sub-regions, the correlation between FH and medianPO2 is low (R(2) = 0.37), while the correlation between FH/P and median PO2 is high (R(2) = 0.99). The ROC analysis showed that hypoxic regions can be identified using FH/P with a higher diagnostic accuracy (YI = sensitivity+ specificity-1 = 1.0), than using FH alone (YI = 0.83). Both FMISO parameters are moderately effective in identifying hypoxia on the microscopic length scale (YI = 0.63 and 0.60).

A combination of two FMISO PET scans acquired 0-15 min and four hours pi may yield an accurate measure of the medianPO2 in a voxel (FH/P). This measure is comparable to averaged oxygen probe measurements and has the advantage of covering the entire tumor volume. Therefore, it may improve the prediction of radiotherapy outcome and facilitate individualized dose prescriptions.