Methodological Improvements Allow the Use of Fractal Geometry as a Prognostic Test in Prostate Cancer
GS Guandalini, V Macias, AK Balla. Armed Forces Institute of Pathology, Washington, DC; University of Illinois at Chicago, Chicago, IL
Background: Fractal dimension (FD) is a relatively new mathematical concept that quantifies irregularity on a given geometric structure. When applied to anatomic pathology, it measures image roughness or complexity. Several approaches have been reported using this technique on cancer histopathology, but so far fractal geometry has not been successfully used as a prognostic marker. Here we examined some crucial methodological improvements that allowed FD to discriminate recurrence from non-recurrence cases after radical prostatectomy.
Design: The source of tissues was the Cooperative Prostate Cancer Tissue Resource (CPCTR; www.prostatetissues.org). We obtained digital images (800 x 600 pixels) of radical prostatectomy H&E slides from 35 subjects with PSA recurrence, matched for race, age, Gleason grading score and pathology TNM classification with 35 patients without recurrence (n=70); a nested case-control design. All non-recurrence subjects had more than 5-years post-prostatectomy follow-up. Snapshots were taken under 20X, 40X, 100X and 200X to find out the effect of magnification to the overall FD. In a novel approach, we applied digitally automated color, brightness and contrast correction (Microsoft Office Picture Manager). FD was obtained performing the box counting method for different scaling windows (ImageJ, NIH freeware). Wilcoxon signed-rank test was performed to compare the paired groups.
Results: No statistically significant difference was observed for images without color, brightness and contrast correction. Significantly increased difference was observed with increasing image correction steps: With five correction steps applied on 40X images, PSA recurrence cases had lower FD than their non-recurrence matches in 26 of the 35 pairs (p = 0.004). Statistically significant results were also observed for 1 and 3 steps corrections at several scaling windows, but only with 40X and 100X magnification.
Conclusions: FD is lower for patients with prostate cancer recurrence when compared to their corresponding disease free matches. This difference suggests that FD might be useful as a prognostic test in prostate cancer; however, enhancing slides color, choosing the proper magnification and using the correct scaling window are decisive procedures to obtain good results.
Monday, March 22, 2010 1:00 PM
Poster Session II # 250, Monday Afternoon