Quantitative Evaluation of Histologic Grade of Breast Cancer Using Digital Image Analysis
Mark C Lloyd, Joseph O Johnson, Jarret House, Evita Henderson, David Tacha, Marilyn M Bui. H. Lee Moffitt Cancer Center, Tampa, FL; University of South Florida, Tampa, FL; Biocare Medical, Concord, CA; Moffitt Cancer Center, Tampa, FL
Background: The Nottingham combined histologic grading system is recommended to grade invasive breast carcinoma. This semi-quantitative evaluation is read by a pathologist using a standard formula. Intra-pathologist variability has been reported. Digital pathology and algorithm development offers the ability to expand the pathologist's toolbox in order to enhance histological diagnosis. This study will investigate the feasibility and reproducibility of quantitative histologic grading of breast cancer.
Design: Twenty-five retrospective cases of ductal and lobular carcinomas of various Nottingham grades were randomly selected. H&E slides were blinded and digitized using a slide scanning instrument. Two pathologists diagnosed and scored each case. Data recorded included Nottingham score for tubule formation, nuclear pleomorphism and mitotic count, combined Nottingham score and grade. Simultaneously, the digital images were used to create a customized image analysis algorithm by using the digital slide to isolate tumor regions via pattern recognition and to score the patterns for each Nottingham scores. The computationally derived scores were compared back to the pathologist's scores. Furthermore, five cases were blindly scored in duplicate to investigate intra-pathologist and intra-image analysis variability.
Results: There was heterogeneity of scoring between pathologists (12/25 cases). The inconsistencies mostly seen in GII cases and was <2 Nottingham scores. Although each pathologist scored 5 cases in duplicate in the same sitting, variability was observed (3/10), one of which resulted in a grade change from II to III. The computer, however, scored all samples with less heterogeneity (<1 Nottingham score) and was 100% consistent in duplicate samples. Quantitative PHH3 based mitotic score indicates a difference between Nottingham grades [GIII 187 mitotic figures in 5mm2 vs 9 in GI].
Conclusions: The development of an image analysis algorithm to quantitatively evaluate histologic grade of breast cancer is feasible and reproducible. This algorithm is not intended to replace the expertise of the pathologist, but to be used as an adjunct tool to enhance the diagnosis. Additionally, PHH3 is being used to molecularly identify mitotic figures. Therefore with further evaluation, more quantifiable scoring methods of mitosis are anticipated to upgrade Nottingham score.
Tuesday, March 20, 2012 1:00 PM
Poster Session IV # 218, Tuesday Afternoon