Umay Kulsoom, Malika Bendechache and Frank G. Glavin, University of Galway, Ireland

Pathology reports are essential documents physicians use to establish a diagnosis and formulate a treatment plan for a specific health condition or disease. The significance of these reports is particularly pronounced in the context of cancer. The accurate classification of these reports is essential for optimising clinical decision-making, ensuring timely interventions, and maintaining high-quality patient care. In this work, we present two key contributions to improve the classification of pathology reports. First, we fine-tuned the Bio+Clinical BERT-based model for a multiclass classification approach that accurately distinguishes between 32 cancer tissues. Second, we have integrated explainability by using LIME to examine the interpretability of the BERT-based model's decisions and identified the domain-specific features that influence the classification results. We have demonstrated that high-performance transformer models can maintain transparency in clinical settings. Our interpretable framework enables pathologists to assess model outputs against established diagnostic criteria, facilitating the responsible integration of clinical language processing systems into clinical workflows.

BERT, NLP, Pathology Reports, Interpretability, Text Classification, LIME.