Quantitative Nuclear Grading: An Objective, Artificial Intelligence-Facilitated Foundation for Grading Noninvasive Papillary Urothelial Carcinoma.

In nonmuscle invasive bladder cancer, grade drives important treatment and management decisions. However, grading is complex and qualitative, and it has considerable interobserver and intraobserver variability. Previous literature showed that nuclear features quantitatively differ between bladder cancer grades, but these studies were limited in size and scope. In this study, we aimed to measure morphometric features relevant to grading criteria and build simplified classification models that objectively distinguish between the grades of noninvasive papillary urothelial carcinoma (NPUC). We analyzed 516 low-grade and 125 high-grade 1.0-mm diameter image samples from a cohort of 371 NPUC cases. All images underwent World Health Organization/International Society of Urological Pathology 2004 consensus pathologist grading at our institution that was subsequently validated by expert genitourinary pathologists from 2 additional institutions. Automated software segmented the tissue regions and measured the nuclear features of size, shape, and mitotic rate for millions of nuclei. Then, we analyzed differences between grades and constructed classification models, which had accuracies up to 88% and areas under the curve as high as 0.94. Variation in the nuclear area was the best univariate discriminator and was prioritized, along with the mitotic index, in the top-performing classifiers. Adding shape-related variables improved accuracy further. These findings indicate that nuclear morphometry and automated mitotic figure counts can be used to objectively differentiate between grades of NPUC. Future efforts will adapt the workflow to whole slides and tune grading thresholds to best reflect time to recurrence and progression. Defining these essential quantitative elements of grading has the potential to revolutionize pathologic assessment and provide a starting point from which to improve the prognostic utility of grade.

Immunohistochemical Assays for Bladder Cancer Molecular Subtyping: Optimizing Parsimony and Performance of Lund Taxonomy Classifiers.

Transcriptomic and proteomic profiling classify bladder cancers into luminal and basal molecular subtypes, with controversial prognostic and predictive associations. The complexity of published subtyping algorithms is a major impediment to understanding their biology and validating or refuting their clinical use. Here, we optimize and validate compact algorithms based on the Lund taxonomy, which separates luminal subtypes into urothelial-like (Uro) and genomically unstable (GU). We characterized immunohistochemical expression data from two muscle-invasive bladder cancer cohorts (n=193, n=76) and developed efficient decision tree subtyping models using 4-fold cross-validation. We demonstrated that a published algorithm using routine assays (GATA3, KRT5, p16) classified basal/luminal subtypes and basal/Uro/GU subtypes with 86%-95% and 67%-86% accuracies, respectively. KRT14 and RB1 are less frequently used in pathology practice but achieved the simplest, most accurate models for basal/luminal and basal/Uro/GU discrimination, with 93%-96% and 85%-86% accuracies, respectively. More complex models with up to eight antibodies performed no better than simpler two- or three-antibody models. We conclude that simple immunohistochemistry classifiers can accurately identify luminal (Uro, GU) and basal subtypes and are appealing options for clinical implementation.