Metastatic prostate cancer can undergo lineage plasticity and evolve into neuroendocrine prostate cancer (NEPC), an aggressive phenotype associated with poor clinical outcomes and therapeutic resistance. The diagnosis of NEPC remains challenging and generally requires additional pathologic evaluation, including specialized immunohistochemical staining.
This study developed and validated NEURAL-PC, an artificial intelligence–based framework that analyzes routinely available histopathology images to identify NEPC without requiring additional stains. By integrating deep learning with interpretable cell-level morphologic features, NEURAL-PC accurately distinguished NEPC across multiple independent cohorts, including metastatic biopsy specimens and samples from a prospective clinical trial. The model also demonstrated prognostic value. These findings support NEURAL-PC as a potentially scalable approach for identifying NEPC from routine histopathology and facilitating earlier recognition and clinical management of this aggressive prostate cancer phenotype.
Read pdf