Automated Detection of Clinical Entities in Lung and Breast Cancer Reports Using NLP Techniques

TL;DR

This work tackles the bottleneck of manual extraction from cancer clinical reports by applying a Named Entity Recognition (NER) approach to Spanish-language EHRs. It fine-tunes a RoBERTa-based model (bsc-bio-ehr-en3) with a manually annotated corpus (200 breast, 400 lung reports) and GMV's uQuery pre-processing to map entities to standard formats. The study demonstrates strong performance for common entities like MET and PAT, and shows that pre-processing substantially boosts accuracy, though underrepresented labels such as EVOL remain challenging. The approach, validated on both breast-only and full datasets (600 reports), indicates high potential for scalable real-world extraction of structured cancer-related data, enabling better real-world evidence and clinical decision support. These results have practical implications for accelerating data-driven cancer research and improving consistency in EHR-derived datasets.

Abstract

Research projects, including those focused on cancer, rely on the manual extraction of information from clinical reports. This process is time-consuming and prone to errors, limiting the efficiency of data-driven approaches in healthcare. To address these challenges, Natural Language Processing (NLP) offers an alternative for automating the extraction of relevant data from electronic health records (EHRs). In this study, we focus on lung and breast cancer due to their high incidence and the significant impact they have on public health. Early detection and effective data management in both types of cancer are crucial for improving patient outcomes. To enhance the accuracy and efficiency of data extraction, we utilized GMV's NLP tool uQuery, which excels at identifying relevant entities in clinical texts and converting them into standardized formats such as SNOMED and OMOP. uQuery not only detects and classifies entities but also associates them with contextual information, including negated entities, temporal aspects, and patient-related details. In this work, we explore the use of NLP techniques, specifically Named Entity Recognition (NER), to automatically identify and extract key clinical information from EHRs related to these two cancers. A dataset from Health Research Institute Hospital La Fe (IIS La Fe), comprising 200 annotated breast cancer and 400 lung cancer reports, was used, with eight clinical entities manually labeled using the Doccano platform. To perform NER, we fine-tuned the bsc-bio-ehr-en3 model, a RoBERTa-based biomedical linguistic model pre-trained in Spanish. Fine-tuning was performed using the Transformers architecture, enabling accurate recognition of clinical entities in these cancer types. Our results demonstrate strong overall performance, particularly in identifying entities like MET and PAT, although challenges remain with less frequent entities like EVOL.

Figures (5)

• Figure 1: Proposed method pipeline, illustrating the stages of the process, from dataset preparation, fine-tuning of the cancer-specific NER model, to the application of the model to pseudonymized clinical records extracted from the Big Data Platform of IIS La Fe and the detection of entities within them.
• Figure 2: Entity extraction from raw text
• Figure 3: Entity extraction from pre-processed text
• Figure 4: (a) Distribution of Hits and Misses in the Validation Set. (b) Comparison of Model-Detected Entities with Expert-Annotated Entities: EVOL (Evolution), FACTR (Risk Factors), MET (Method of Diagnosis), NO_LABEL (Non annotated entities), PAT (Pathology), SINT (Symptomatology), and TTO (Treatment).
• Figure 5: (a) Distribution of Hits and Misses in the Validation Set. (b) Comparison of Model-Detected Entities with Expert-Annotated Entities: EVOL (Evolution), FACTR (Risk Factors), ANTPERSON (Personal History, specific to lung cancer), MUTAC (Genetic Mutations, specific to lung cancer),MET (Method of Diagnosis), PAT (Pathology), SINT (Symptomatology), and TTO (Treatment).