Search

The use of artificial intelligence (AI) within pathology and health care has advanced extensively. We have accordingly witnessed an increased adoption of various AI tools that are transforming our approach to clinical decision support, personalized medicine, predictive analytics, automation, and discovery. The familiar and more reliable AI tools that have been incorporated within health care thus far fall mostly under the nongenerative AI domain, which includes supervised and unsupervised machine learning (ML) techniques. This review article explores how such nongenerative AI methods, rooted in traditional rules-based systems, enhance diagnostic accuracy, efficiency, and consistency within medicine. Key concepts and the application of supervised learning models (ie, classification and regression) such as decision trees, support vector machines, linear and logistic regression, K-nearest neighbor, and neural networks are explained along with the newer landscape of neural network-based nongenerative foundation models. Unsupervised learning techniques, including clustering, dimensionality reduction, and anomaly detection, are also discussed for their roles in uncovering novel disease subtypes or identifying outliers. Technical details related to the application of nongenerative AI algorithms for analyzing whole slide images are also highlighted. The performance, explainability, and reliability of nongenerative AI models essential for clinical decision-making is also reviewed, as well as challenges related to data quality, model interpretability, and risk of data drift. An understanding of which AI-ML models to employ and which shortcomings need to be addressed is imperative to safely and efficiently leverage, integrate, and monitor these traditional AI tools in clinical practice and research., (Copyright © 2024 The Authors. Published by Elsevier Inc. All rights reserved.)

The HercepTest was approved 20+ years ago as the companion diagnostic test for trastuzumab in human epidermal growth factor 2 (HER2) or ERBB2 gene-amplified/overexpressing breast cancers. Subsequent HER2 immunohistochemistry (IHC) assays followed, including the now most common Ventana 4B5 assay. Although this IHC assay has become the clinical standard, its reliability, reproducibility, and accuracy have largely been approved and accepted on the basis of concordance among small numbers of pathologists without validation in a real-world setting. In this study, we evaluated the concordance and interrater reliability of scoring HER2 IHC in 170 breast cancer biopsies by 18 breast cancer-specialized pathologists from 15 institutions. We used the Observers Needed to Evaluate Subjective Tests method to determine the plateau of concordance and the minimum number of pathologists needed to estimate interrater agreement values for large numbers of raters, as seen in the real-world setting. We report substantial discordance within the intermediate categories (<1% agreement for 1+ and 3.6% agreement for 2+) in the 4-category HER2 IHC scoring system. The discordance within the IHC 0 cases is also substantial with an overall percent agreement (OPA) of only 25% and poor interrater reliability metrics (0.49 Fleiss' kappa, 0.55 intraclass correlation coefficient). This discordance can be partially reduced by using a 3-category system (28.8% vs 46.5% OPA for 4-category and 3-category scoring systems, respectively). Observers Needed to Evaluate Subjective Tests plots suggest that the OPA for the task of determining a HER2 IHC score 0 from not 0 plateaus statistically around 59.4% at 10 raters. Conversely, at the task of scoring HER2 IHC as 3+ or not 3+ pathologists' concordance was much higher with an OPA that plateaus at 87.1% with 6 raters. This suggests that legacy HER2 IHC remains valuable for finding the patients in whom the ERBB2 gene is amplified but unacceptably discordant in assigning HER2-low or HER2-negative status for the emerging HER2-low therapies., (Copyright © 2022 United States & Canadian Academy of Pathology. Published by Elsevier Inc. All rights reserved.)

Invasive lobular carcinoma with extracellular mucin (ILCEM) is a rare histologic subtype of breast cancer. Little is known about the pathologic or genomic signatures that distinguish ILCEM from classic invasive lobular carcinoma (ILC) or mucinous carcinoma. We studied 17 breast cancers with lobular morphology and extracellular mucin. Thirteen tumors with sufficient tissue for DNA extraction were analyzed by a next generation sequencing (NGS) assay that interrogates 447 genes for mutations and copy number variations (CNVs). Median patient age was 66 yrs (range: 31-77 yrs). Sixteen patients presented with masses, 7 of which were >2 cm. Seven patients had lymph node metastases. The cases of ILCEM were moderately (n = 13) or poorly differentiated (n = 4), frequently exhibiting variant morphology that has not been previously described or emphasized, including grade 3 nuclei (n = 11), diffuse signet ring cells (n = 10), solid growth (n = 4), tumor necrosis (n = 3) or apocrine features (n = 2). All tumors showed absent or reduced membranous E-cadherin expression. Concurrent lobular carcinoma in situ (LCIS) was seen in 11/17 cases, 1 of which was a striking example of signet ring cell LCIS with extracellular mucin. Receptor profiles were ER+/HER2- (n = 15) and ER+/HER2+ (n = 2). With a median follow-up of 83.5 months (range: 3-171 months) in 12 patients with available information, 8 patients had recurrences resulting in 4 cancer-related deaths. The most common CNVs were 16q loss (n = 11) and 1q gain (n = 9). CDH1 gene-level alterations were detected in all but one case, including frameshift (n = 7), nonsense (n = 2), and donor splice site (n = 1) mutations and indels (n = 2). Recurrent mutations were also seen in PIK3CA (n = 3), POLQ (n = 3), TP53 (n = 3), ERBB3 (n = 3), ERBB2 (n = 2), and RUNX1 (n = 2). Genes with recurrent amplifications included GATA3 (n = 4), FOXA1 (n = 3), CCND1 (n = 2). Our data highlights ILCEM as a distinct variant of ILC that often presents with higher-grade and variant morphologic features and is associated with an aggressive clinical course. NGS data support an overall lobular-type molecular profile and reveal potentially targetable alterations in a subset of cases with recurrence., (© 2022. The Author(s), under exclusive licence to United States & Canadian Academy of Pathology.)

Pleomorphic LCIS (P-LCIS) and florid LCIS (F-LCIS) are morphologic variants distinguished from classic LCIS by marked nuclear pleomorphism and/or an expansile growth pattern with or without necrosis. Given the rarity of these LCIS variants, little data exist regarding their molecular pathogenesis, natural history, and optimal management. The purpose of this study was to genomically profile LCIS variants to gain further insight into their biology. Nineteen cases of pure LCIS variants (17 P-LCIS, 2 F-LCIS) diagnosed on core needle biopsy at our institution from 2006 to 2017 were included, five of which were upgraded to invasive cancer at excision. Macrodissected lesions were analyzed by a hybrid-capture next generation sequencing assay that surveyed exonic sequences of 447 genes for mutations and copy number variations (CNVs) and 191 regions across 60 genes for structural rearrangements. LCIS variants were all confirmed as E-cadherin negative by immunohistochemistry. Receptor profiles among the 17 P-LCIS cases included HR+/HER2- (nine cases), HR+/HER2+ (three cases), HR-/HER2+ (two cases), and HR-/HER2- (three cases). The two F-LCIS cases were HR+/HER2- and HR+/HER2+. All LCIS variants had genetic alterations consistent with a lobular phenotype including 1q gain (16 cases), 16q loss (18 cases), and CDH1 mutations (18 cases). Highly recurrent ERBB2 alterations were noted including mutations (13 cases) and amplifications (six cases). Other significant alterations included mutations in PIK3CA (six cases), RUNX1 (four cases), ERBB3 (four cases), and CBFB (three cases), as well as amplification of CCND1 (five cases). A TP53 mutation was identified in one case of HR-/HER2+ P-LCIS with signet ring cell features that lacked 1q gain and 16q loss. P-LCIS and F-LCIS contain genetic alterations characteristic of lobular neoplasia; however, these LCIS variants are distinguished from classical LCIS reported in the literature by their highly recurrent ERBB2 alterations.