A novel method for analyzing DSCE-images with an application to tumor grading.

OBJECTIVES: (a) The development of a novel analysis method, named Dynamic pixel intensity Histogram Analysis (DHA) allowing for pixel intensity-histogram-model-parameter fitting of arbitrary-shaped regions defined in dynamic-susceptibility-contrast-enhanced (DSCE) difference MR-image time-series, and (b) its prospective application and evaluation for glioma grading. MATERIALS AND METHODS: For each difference-image, pixel intensity histograms of arbitrary-shaped ROIs were computed and fitted using the Levenberg-Marquardt algorithm. Time-dependent histogram center-position- and width-parameters are computed during bolus-passage. The method was applied to 25 patients with low and high grade gliomas. RESULTS: During bolus outflow-time, histogram-center-position-parameter and histogram-width-parameter reach highest significance levels and discriminate gliomas of different grades. The histogram center-position-parameter discriminated grade-II from grade-III, grade-II from grade-IV but not grade-III from grade-IV. The observed histogram width-parameters discriminated grade-II from grade-III (P < 0.00022), grade-II from grade-IV (P <8.3 10), and grade-III from grade-IV (P < 0.00063). CONCLUSIONS: DHA is a easy-to-use method for glioma grading; the histogram width parameter is best indicator for histologic grade.