A Quantitative Method for Iron Histochemistry
David P Ng, Lauren C Hyde, Norman B Levy, Deborah L Ornstein. Dartmouth-Hitchcock Medical Center, Lebanon, NH
Background: Currently, iron quantitation in histologic sections relies on discrete grades derived from manual examination of slides stained for ferric iron. In the bone marrow, several grading systems are used, ranging from a simple 3 grade score (absent/present/increased) to the more complex Gale's score with 7 grades. Several studies have suggested a high degree of interobserver variability in the application of these scoring schemes as well as inconsistency in the clinical use of these methods, suggesting a need for more reproducible and consistent methods to quantitate iron in histologic sections. Here, we report a new method to quantitate storage iron in histologic sections using open source image analysis tools.
Design: Bone marrow core biopsy sections were stained with Perl's iron and a nuclear fast red counterstain. Ten representative fields of each core were evaluated from whole slide scanned images to eliminate camera field variability. Using ImageJ, the images were separated into RGB color channels, and the red and blue channels were thresholded so that cellular (red) pixels and ferric iron (blue) pixels were selected. A ratio of number of ferric iron pixels to cellular pixels was calculated for each image and an overall average was calculated for each specimen to derive a quantitative Iron Score (qIS). Twenty sequential cores were graded by two hematopathologists using Gale's scoring system while qIS were obtained by the method above and the results compared. Interobserver variability of the field selection was analyzed with representative images from an additional 154 cores evaluated by trained and untrained independent observers for a total of 3080 images. Linear regression between qIS was performed.
Results: Using Gale's iron scoring system, the two hematopathologists showed good correlation with the log normalized qIS (R2=0.696 and 0.663) similar to Gale's original results using radio-chemical iron analysis. Field selection by trained and untrained observers yielded excellent correlation for qIS (ρ=1.006, R2=0.9661).
Conclusions: These results suggest that our method is useful for quantitating storage iron in bone marrow biopsy specimens and correlates well with older semi-quantitative methods. Additionally, this technique appears to be very robust and tolerant of the method by which fields are selected for analysis. This suggests that untrained observers may be trained to select fields for analysis. The use of whole slide scanning and ImageJ makes this method highly reproducible with little inter-run and inter-observer variability and can be readily applied in any lab with a camera and computer.
Wednesday, March 6, 2013 9:30 AM
Poster Session V # 290, Wednesday Morning