Machine Learning Models Improve the Diagnostic Yield of Peripheral Blood Flow Cytometry.

OBJECTIVES: Peripheral blood flow cytometry (PBFC) is useful for evaluating circulating hematologic malignancies (HM) but has limited diagnostic value for screening. We used machine learning to evaluate whether clinical history and CBC/differential parameters could improve PBFC utilization. METHODS: PBFC cases with concurrent/recent CBC/differential were split into training (n = 626) and test (n = 159) cohorts. We classified PBFC results with abnormal blast/lymphoid populations as positive and used two models to predict results. RESULTS: Positive PBFC results were seen in 58% and 21% of training cases with and without prior HM (P < .001). % neutrophils, absolute lymphocyte count, and % blasts/other cells differed significantly between positive and negative PBFC groups (areas under the curve [AUC] > 0.7). Among test cases, a decision tree model achieved 98% sensitivity and 65% specificity (AUC = 0.906). A logistic regression model achieved 100% sensitivity and 54% specificity (AUC = 0.919). CONCLUSIONS: We outline machine learning-based triaging strategies to decrease unnecessary utilization of PBFC by 35% to 40%.

Morphologists overestimate the nuclear-to-cytoplasmic ratio.

BACKGROUND: The Paris System for Reporting Urinary Cytology (TPS) has defined nuclear-to-cytoplasmic (N:C) ratio cutoff values for several of its risk-stratified diagnostic categories. However, because pathologists are not trained to recognize strict N:C ratio cutoff values, a previously designed survey was used to determine whether pathologists could accurately identify N:C ratios according to TPS standards. METHODS: Participants were instructed to estimate the N:C ratio of ideal (line drawing) and real (cell photograph) images presented via an online survey. Actual N:C ratios ranged from 0.3 to 0.8, and 3 answer choices were available: < 0.5, ≥ 0.5 and <0.7, and ≥0.7. The resulting data were analyzed to determine the accuracy and performance of the subgroups. RESULTS: A total of 137 individuals completed the survey. Approximately 24.1% were cytopathologists, 18.2% were pathologists without formal cytopathology training, 18.2% were cytotechnologists, 24.1% were pathology residents, and 15.3% were nonmorphologists. Overall, 70.0%, 67.6%, and 93.3% of responses, respectively, were correct for images with an N:C ratio of < 0.5, ≥0.5 and < 0.7, and ≥0.7. For images with an actual N:C ratio < 0.5 and ≥0.5 and < 0.7, 30.0% and 25.0% of responses, respectively, overestimated the N:C ratio. Furthermore, for images with an N:C ratio of 0.4 and 0.6, > 40.0% of responses overestimated the N:C ratio. As a whole, morphologists were significantly more accurate than nonmorphologists (P = .030). CONCLUSIONS: Morphologists tended to overestimate the N:C ratio, particularly at ratios close to TPS-recommended cutoff values. Additional training regarding N:C ratio estimation may help pathologists to adapt to this new system. Cancer Cytopathol 2016;124:669-77. © 2016 American Cancer Society.