Natural language processing in Clinical Medicine

Natural Language Processing (NLP) in clinical medicine is attracting increasing attention, especially given the recent advancements in generative AI. The emergence of Generative Pretrained Transformers (GPT) - a type of Pretrained Language Models (PLM) - is challenging the role of traditional rule based and machine learning approaches in Clinical NLP. Giant Tech Companies like OpenAI, Microsoft, and Google have made PLMs commercially accessible AI products. Furthermore, open source and cloud-based programming advancements also  create the possibility of training the next generation of NLP analysts focusing on clinical applications. While there is potential for PLMs/Generative AI in clinical applications, numerous issues stemming from privacy, adaptability, and comprehensive evaluation should be carefully investigated before applying them to medical usage.

At present, hospital Electronic Health Records (EHR) still rely heavily on narrative free text to capture patient medical history. Clinical NLP has potential to liberate free text collection usage to enhance clinical outcomes and ensure patient safety and performance monitoring. Furthermore, understanding how NLP-structured clinical data can be combined with other routinely collected EHR data is essential for improving treatment and prognosis outcomes.

In recognition of the increasingly significant role that clinical NLP will play in the next generation of digital advancements and in enhancing seamless interaction in the future, this Collection welcomes NLP innovations to improving medical and population health outcomes. Submissions should focus on evidence-based computational linguistics approaches and applications for health and adhere to providing impact on digital medicine.

Subtopics include (but are not limited to):

• Innovative approaches for accurately extracting narrative information from hospital notes, including annotation schema development, transformers, zero-shot/few-shot learning regime, biomedical vocabularies. Areas include, but are not limited to, patient safety, deidentification, infection control, cancer diagnosis and prognosis.
• Language modeling strategies to address challenges arising from privacy in medicine, linguistic and cultural diversity, and socioeconomic disparities.
• NLP-integrated digital health applications applied into administrative and clinical notes, discussion threads between clinicians and patients or patient-reported narratives, aiming to enhance clinical process and medical/population outcomes.
• Analysis of social trends and patient sentiment to understand health behaviors and outcomes, particularly in the context of pandemics and universal health coverage.
• Novel approaches and initiatives in clinical NLP education to empower patients and build future capability.

This Collection supports and amplifies research related to SDG 3: Good health and well-being.