Enumeration of CD34 Positive Blasts in Bone Marrow Biopsy Specimens by Digital Image Analysis
Jeffrey M Barroso, Zixuan Wang, Kristen Velazquez, Alina Dulau-Florea, Jerald Z Gong. Thomas Jefferson University Hospital, Philadelphia, PA
Background: Accurate enumeration of blasts is an essential component in the diagnosis and subtyping of acute and chronic leukemias. Manual count on bone marrow aspirate smears is currently still the preferred method for blast enumeration, but flow cytometry (FC) and immunohistochemistry (IHC) are often used as alternatives, especially when aspirate specimens are suboptimal. CD34 expression by IHC has been widely accepted as a tool in the assessment of blasts, however, the microscopic estimate of CD34 positive cells is semiquantitative and often insufficient for the initial diagnosis and classification of myeloid leukemias. Here, we show that digital image analysis (DIA) can provide more information than estimate alone.
Design: Fifty-four cases of CD34 immunostained bone marrow biopsies were retrieved from pathology files and scanned at 200x magnification with an automated computer scanner (Aperio ScanScope). Using separate membrane and nuclear algorithms to determine positive and negative cells, CD34 expression from three representative areas, each with at least 500 cells, was analyzed. CD34 positivity was then compared with CD34 blast count by microscopic estimate, manual blast count on bone marrow aspirate smears, and blast count by FC analysis. Statistical analyses were performed using Pearson correlation and linear regression analysis.
Results: CD34 positive cells ranged from 1-37% (average 16%) by DIA and 3-90% (average 24%) by microscopic estimate. DIA provided quantitative results with continuous data, while microscopic estimate was semiquantitative with a 5 percent point increment. Percentage of CD34 positive cells by DIA strongly correlated with the microscopic estimate when blasts were <35% (r=0.71), but poorly correlated when blasts were >35% due to algorithmic underestimation. Weaker correlation was observed between DIA and manual differential count (r=0.49) and between DIA and FC (r=0.40). This was likely caused by hemodilution, partial expression of CD34, and technical factors associated with aspirate smears and FC analysis.
Conclusions: DIA for enumeration of CD34 positive blasts in myeloid neoplasms was comparable with microscopic estimate in cases with a lower blast percentage (<35%). It provided more information with quantitative data and may have potential in assisting classification of myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) when bone marrow aspirate is suboptimal. Further studies to assess DIA on specific types of MDS and AML are needed to verify our results in the subtyping of these diseases.
Wednesday, March 21, 2012 1:00 PM
Poster Session VI # 242, Wednesday Afternoon