Masked pre-training of transformers for histology image analysis.

In digital pathology, whole-slide images (WSIs) are widely used for applications such as cancer diagnosis and prognosis prediction. Vision transformer (ViT) models have recently emerged as a promising method for encoding large regions of WSIs while preserving spatial relationships among patches. However, due to the large number of model parameters and limited labeled data, applying transformer models to WSIs remains challenging.

Artificial intelligence in digital pathology of cutaneous lymphomas: A review of the current state and future perspectives.

Primary cutaneous lymphomas (CLs) represent a heterogeneous group of T-cell lymphomas and B-cell lymphomas that present in the skin without evidence of extracutaneous involvement at time of diagnosis. CLs are largely distinct from their systemic counterparts in clinical presentation, histopathology, and biological behavior and, therefore, require different therapeutic management. Additional diagnostic burden is added by the fact that several benign inflammatory dermatoses mimic CL subtypes, requiring clinicopathological correlation for definitive diagnosis.

Prevalence, clinical and molecular characteristics of early stage EGFR-mutated lung cancer in a real-life West-European cohort: Implications for adjuvant therapy.

Objectives: The landmark ADAURA study recently demonstrated a significant disease-free survival benefit of adjuvant osimertinib in patients with resected EGFR-mutated lung adenocarcinoma. However, data on prevalence rates and stage distribution of EGFR mutations in non-small cell lung cancer in Western populations are limited since upfront EGFR testing in early stage lung adenocarcinoma is not common practice. Here, we present a unique, real-world, unselected cohort of lung adenocarcinoma to aid in providing a rationale for routine testing of early stage lung cancers for EGFR mutations in the West-European population.

The sensitivity of pan-TRK immunohistochemistry in solid tumours: A meta-analysis.

Introduction: Since the approval of neurotrophic tropomyosin receptor kinase (NTRK) tyrosine kinase inhibitors for fist-line advanced stage pan-cancer therapy, pathologists and molecular biologists have been facing a complex question: how should the large volume of specimens be screened for NTRK fusions? Immunohistochemistry is fast and cheap, but the sensitivity compared to RNA NGS is unclear.

Methods: We performed RNA-based next-generation sequencing on 1,329 cases and stained 24 NTRK-rearranged cases immunohistochemically with pan-TRK (ERP17341). Additionally, we performed a meta-analysis of the literature.

The impact of a pathologist's personality on the interobserver variability and diagnostic accuracy of predictive PD-L1 immunohistochemistry in lung cancer.

Objectives: Programmed death-ligand 1 (PD-L1) is the only approved predictive biomarker for immunotherapy in non-small cell lung cancer (NSCLC). However, predictive PD-L1 immunohistochemistry is subject to interobserver variability. We hypothesized that a pathologist's personality influences the interobserver variability and diagnostic accuracy of PD-L1 immunoscoring.

Real-World Approach for Molecular Analysis of Acquired EGFR Tyrosine Kinase Inhibitor Resistance Mechanisms in NSCLC.

Introduction: With the approval of first-line osimertinib treatment in stage IV EGFR-mutated NSCLC, detection of resistance mechanisms will become increasingly important-and complex. Clear guidelines for analyses of these resistance mechanisms are currently lacking. Here, we provide our recommendations for optimal molecular diagnostics in the post-EGFR tyrosine kinase inhibitor (TKI) resistance setting.

Development and validation of a supervised deep learning algorithm for automated whole-slide programmed death-ligand 1 tumour proportion score assessment in non-small cell lung cancer.

Aims: Immunohistochemical programmed death-ligand 1 (PD-L1) staining to predict responsiveness to immunotherapy in patients with advanced non-small cell lung cancer (NSCLC) has several drawbacks: a robust gold standard is lacking, and there is substantial interobserver and intraobserver variance, with up to 20% discordance around cutoff points. The aim of this study was to develop a new deep learning-based PD-L1 tumour proportion score (TPS) algorithm, trained and validated on a routine diagnostic dataset of digitised PD-L1 (22C3, laboratory-developed test)-stained samples.

Methods And Results: We designed a fully supervised deep learning algorithm for whole-slide PD-L1 assessment, consisting of four sequential convolutional neural networks (CNNs), using aiforia create software.

Optimizing Mutation and Fusion Detection in NSCLC by Sequential DNA and RNA Sequencing.

Introduction: Frequently, patients with locally advanced or metastatic NSCLC are screened for mutations and fusions. In most laboratories, molecular workup includes a multitude of tests: immunohistochemistry (ALK, ROS1, and programmed death-ligand 1 testing), DNA sequencing, in situ hybridization for fusion, and amplification detection. With the fast-emerging new drugs targeting specific fusions and exon-skipping events, this procedure harbors a growing risk of tissue exhaustion.