[The model transferability of AI in digital pathology : Potential and reality]. / Die Modelltransferierbarkeit von KI in der digitalen Pathologie : Potenzial und Realität.

OBJECTIVE: Artificial intelligence (AI) holds the potential to make significant advancements in pathology. However, its actual implementation and certification for practical use are currently limited, often due to challenges related to model transferability. In this context, we investigate the factors influencing transferability and present methods aimed at enhancing the utilization of AI algorithms in pathology. MATERIALS AND METHODS: Various convolutional neural networks (CNNs) and vision transformers (ViTs) were trained using datasets from two institutions, along with the publicly available TCGA-MIBC dataset. These networks conducted predictions in urothelial tissue and intrahepatic cholangiocarcinoma (iCCA). The objective was to illustrate the impact of stain normalization, the influence of various artifacts during both training and testing, as well as the effects of the NoisyEnsemble method. RESULTS: We were able to demonstrate that stain normalization of slides from different institutions has a significant positive effect on the inter-institutional transferability of CNNs and ViTs (respectively +13% and +10%). In addition, ViTs usually achieve a higher accuracy in the external test (here +1.5%). Similarly, we showcased how artifacts in test data can negatively affect CNN predictions and how incorporating these artifacts during training leads to improvements. Lastly, NoisyEnsembles of CNNs (better than ViTs) were shown to enhance transferability across different tissues and research questions (+7% Bladder, +15% iCCA). DISCUSSION: It is crucial to be aware of the transferability challenge: achieving good performance during development does not necessarily translate to good performance in real-world applications. The inclusion of existing methods to enhance transferability, such as stain normalization and NoisyEnsemble, and their ongoing refinement, is of importance.

[Histomolecular classification of urothelial carcinoma of the urinary bladder : From histological phenotype to genotype and back]. / Histomolekulare Klassifikation des Urothelkarzinoms der Harnblase : Vom histologischen Phänotyp zum Genotyp und zurück.

BACKGROUND: Of all urothelial carcinomas (UCs), 25% are muscle invasive and associated with a 5-year overall survival rate of 50%. Findings regarding the molecular classification of muscle-invasive urothelial carcinomas (MIUCs) have not yet found their way into clinical practice. OBJECTIVES: Prediction of molecular consensus subtypes in MIUCs with artificial intelligence (AI) based on histologic hematoxylin-eosin (HE) sections. METHODS: Pathologic review and annotation of The Cancer Genome Atlas (TCGA) Bladder Cancer (BLCA) Cohort (N = 412) and the Dr. Senckenberg Institute of Pathology (SIP) BLCA Cohort (N = 181). An AI model for the prediction of molecular subtypes based on annotated histomorphology was trained. RESULTS: For a five-fold cross-validation with TCGA cases (N = 274), an internal TCGA test set (N = 18) and an external SIP test set (N = 27), we reached mean area under the receiver operating characteristic curve (AUROC) scores of 0.73, 0.8 and 0.75 for the classification of the used molecular subtypes "luminal", "basal/squamous" and "stroma-rich". By training on correlations to individual molecular subtypes, rather than training on one subtype assignment per case, the AI prediction of subtypes could be significantly improved. DISCUSSION: Follow-up studies with RNA extraction from various areas of AI-predicted molecular heterogeneity may improve molecular classifications and thereby AI algorithms trained on these classifications.