Programme

Chairs: Riza Batista-Navarro (University of Manchester) and Denis Newman-Griffis (University of Sheffield)

• Matúš Falis. Can ChatGPT Generate and Code Discharge Summaries?
• Nastazja Laskowski. Data Transparency and Anonymization when Sharing Clinical Study Reports: An exploration of Natural Language Processing and Statistical Disclosure Control
• Ratchakrit Arreerard. Feasibility of Emotions as Features for Suicide Ideation Detection in Social Media

Panel: Angus Roberts (University of Sheffield), Yonghui Wu (University of Florida), Sarah Markham (patient representative), Anoop Shah (University College London, UCLH)

Chair: Dr Liz Ford (Brighton and Sussex Medical School)

This panel will discuss how patients and members of public can be involved in co-production during the entire clinical NLP application lifecycle: from conception of the idea to the deployment and monitoring of its efficacy and impact. The panellists include:

Niccola Hutchinson-Pascal: Niccola is part of Co-Production Collective, a co-produced community working to support the authentic co-production (not faux-production!) of research, service and policy development. Everyone is welcome. Together we learn, connect, and champion co-production for lasting change. Niccola has worked for a wide variety of organisations across health, well-being and physical activity, from charities, to government related bodies, to large agencies. All of these roles have had a focus on culture change and involved her working closely with the public, patients and local community members. She is passionate about co-production, about all parties communicating on a level playing field, sharing power and decision making and about ensuring organisations are aware of the value gained from this way of working.

Debbie Keatley: Debbie is a cancer survivor who believes in the patient’s right to be an equal partner in treatment and care decisions and in the meaningful involvement of the public and patients in research. Impatient for data resources held to be used to benefit patients, her ambition is for people across the UK to be able to access and understand their own health records and to make informed choices about how their data is used in research, with confidence that regulatory institutions will respect and uphold those decisions. Involved with many organisations who recognise data’s importance in research to improve outcomes for patients, including use MY data, CRUK’s Data Advisory Board and Clinical Research Committee, the office of the National Data Guardian and Northern Ireland Cancer Registry and a former member of HDRUK’s Public Advisory Board, she is actively involved in steering groups of clinical studies across Northern Ireland, the UK and Europe and in the work of Ireland’s All-Island Cancer Research Institute.

Sophie Gibbons: Sophie is a Research Scientist at Akrivia Health, has a core role in empowering Akrivia’s NHS partners to unlock insights more easily from their existing data. Additional to this, Sophie works on: ensuring the research utility and transparency of AI approaches to deriving structured data from the free-text portion of electronic health records (namely, the application and communication of Akrivia’s Natural Language Processing); co-development of approaches to mental health data access with Akrivia’s in-house Patient and Public Involvement (PPI) team; and expansion of Akrivia’s research network such that more high quality research can leverage Akrivia’s valuable data asset. Sophie will talk on Patient involvement in developing steps to access mental health medical data.

The Maudsley Biomedical Research Centre has been developing and applying natural language processing methods to identify a variety of clinical variables in clinical free-text notes. This talk will review the NLP developments, from dictionary and rule-based approaches to training, fine-tuning and evaluation of pre-trained language models.

Chair: Honghan Wu

• James Brandreth, Jennifer Jiang and Anoop Shah. MiADE (Medical information AI Data Extractor): Natural language processing at the point of care (DEMO)
• Jaya Chaturvedi, Diana Shamsutdinova, Felix Zimmer, Sumithra Velupillai, Daniel Stahl, Robert Stewart and Angus Roberts. Sample Size in Natural Language Processing within Healthcare Research
• Sophie Gibbons, Panagiota Kontari, Simon Pillinger, Elizabeth Ford and Ben Fell. Patient and Public Involvement Co-Development of a Route to Record-Level Data Access to Akrivia Health’s Secondary Healthcare Dataset
• Bernadeta Griciūtė, Lifeng Han, Hao Li and Goran Nenadic. Topic Modelling of Swedish Newspaper Articles about Coronavirus: a Case Study using Latent Dirichlet Allocation Method
• Malik Ahmed. Combining Rule-Based Techniques and GPT-4 for Clinical Drug Information Extraction from SmPC Documents: A Natural Language Processing Approach to Developing Accessible and Up-to-Date Drug Databases
• Areej Alhassan, Viktor Schlegel, Monira Aloud, Riza Batista-Navarro and Goran Nenadic. Towards Recognising Discontinuous Named Entities in Clinical Text Using a Seq2Seq Prompt-guided Model
• Sean Farrell, Charlotte Appleton, Peter-John Mäntylä Noble and Noura Al Moubayed. PetBERT: Automated ICD-11 Syndromic Disease Coding for Outbreak Detection in First Opinion Veterinary Electronic Health Records
• Matthew Coole, Paul Rayson, Paul Marshall and Fiona Lobban. iPOF: Improving Peer Online Forums Project Progress
• Tomas Goldsack, Zhihao Zhang, Chenghua Lin and Carolina Scarton. Making Science Simple: Corpora for the Lay Summarisation of Biomedical Literature
• F. Dalla Serra, G. Jacenków, F. Deligianni, J. Dalton and A Q. O’Neil. Improving Image Representations via MoCo Pre-Training for Multimodal CXR Classification
  
• Reem Bin-Hezam and Mark Stevenson. Improving Stopping in Technology Assisted Reviews
  

Chair: Rob Stewart (KCL)

The panel will discuss the practice of producing effective clinical annotation guidelines that both capture clinical intention and provide usable recipes for textual annotation. What are common principles and steps in developing clinical annotation guidelines? What lessons have we learnt so far? How can we make clinical annotation guidelines FAIR (Findabile, Accessibile, Interoperable, and Reusable)?

Panel: Ben Fell (Akrivia Health), Warren Del-Pinto (University of Manchester), Eulàlia Farré-Maduell (Barcelona Supercomputing Center), Imane Guellil (University of Edinburgh)

• Standards for data modelling and representation (OMOP, FHIR)
• Towards guidelines for clinical annotation guidelines
• PPIE
• Clinical NLP governance

• Daphne Chopard, Padraig Corcoran and Irena Spasic. Word Sense Disambiguation of Acronyms in Clinical Narratives
• Arlene Casey, Emma Davidson, Claire Grover, Richard Tobin, Andreas Grivas, Huayu Zhang, Patrick Schrempf, Alison Q. O’Neil, Liam Lee, Michael Walsh, Frey Pellie, Karen Ferguson, Vera Cvero, Honghan Wu, Heather Whalley, Grant Mair, William Whiteley and Beatrice Alex. Understanding performance and reliability of NLP tools: A comparison of four NLP tools predicting stroke phenotypes in radiology reports
• Emma Davidson, Arlene Casey, Claire Grover, Beatrice Alex, Honghan Wu, Archie Campbell, Fionna Chalmers, Mark Adams, Matthew Iveson, Andrew M Macintosh, Emily Ball, Kristiina Rannikmae, Heather Whalley and William Whiteley. The epidemiological characteristics of stroke phenotypes defined with ICD-10 and free-text: a cohort study linked to electronic health records
• Anna-Grace Linton, Vania Dimitrova, Amy Downing, Richard Wagland and Adam Glaser. Weakly Supervised Text Classification on Free Text Comments in Patient-Reported Outcome Measures
• Ghada Alfattni, Niels Peek, Anthony Wilson and Goran Nenadic. Integrating Patients’ Medication Histories from Structured and Unstructured Data
• Siqu Long, Shuang Zhang, Feiqi Cao, Josiah Poon and Soyeon Han. Suicide-NLU: Suicidality Detection by using Joint Intent Classification and Slot Filling

• Maksim Belousov, Vladislav Yotkov. Can Digital Humans transform healthcare? (DEMO from Re:course)
• Thanh-Tung Nguyen, Viktor Schlegel, Abhinav Ramesh Kashyap, Hao Li and Stefan Winkler. Mimic-IV-ICD: A new benchmark for eXtreme MultiLabel Classification
  

• Ceyda Uysal, Gloria Roque, Tarso Franarin, Luke Brydon, Sophie Gibbons and Benjamin Fell. Contextual Classification of Substance Use in Electronic Health Records
• Yang Cui, Lifeng Han and Goran Nenadic. Prompt-based Temporal Classification of Treatment Events from Discharge Summaries
• Oli Delgaram-Nejad, Dawn Archer and Gerasimos Chatzidamianos. Developing the 4SC: a Small, Specialised, Spoken, Schizophrenia corpus
• Arooj Hussain, Haifa Alrdahi, Hendrik Šuvalov, Lifeng Han, Goran Nenadic, Nenadic, Will Dixon and Meghna Jani. M3: Extracting medication and related attributes from outpatient letters
• Zhen Zhu. Insights From Diabetes-Related Food Reviews on an E-Commerce Platform: A Text Analytics Approach
• Ivo Fins, Heather Davies, Sean Farrell and Peter-John Noble. Can ChatGPT help to tackle canine obesity? – Uncovering Body Condition Score measurements and overweight companion animals described in veterinary clinical narratives using ChatGPT. Performance comparison with a regular expression-based tool
• Shuai Niu and Xian Yang. Enhancing Clinical Decision Making with Interpretable Evidence for Personalized Disease Risk Prediction
• Xu Wang, Edward Meinert, Andrea Preston and Shang-Ming Zhou. Identification of Influential Factors in Bladder Cancer: A Co-Designed Study by Utilizing Epidemiology and Machine Learning Framework on Large Electronic Health Records Cohort
• Antanas Kascenas, Nicolas Pugeault, Alison Q. O’Neil. Denoising Autoencoders for Unsupervised Anomaly Detection in Brain MRI

Chair: Prof James Teo (KCL)

This panel will focus on exploring how clinical NLP applications should be evaluated in terms of their fitness for purpose: what steps and metrics are needed in the process of evaluation of NLP software for use with real-world data? How to estimate the quality and utility of the software: is comparison to a gold standard enough? If so, what is the right sample size? If not, what else needs to be evaluated? How to evaluate scalability and architectural fit? How the evaluation outcomes are to be communicated to stakeholders?

Panel: Will Dixon (University of Manchester), Darren Lunn (Medicines and Healthcare products Regulatory Agency), Rob Brisk (Eolas Medical), Arlene Casey (University of Edinburgh)

Chair: Richard Dobson

Chair: Dr Dan Schofield (NHS England)

Industry panel will discuss the opportunities and challenges of generative language models in healthcare, and the roles that industry, academia and NHS would play when these models are trained, evaluated and used to support various activities. How would these models change the landscape of healthcare text analytics? What new challenges such models bring to industry involvement? How will we ensure privacy and fairness, in particular if the models are not tuned on local data? The panellists will be from industry, academia, governance and patient communities.

Panel: Yonghui Wu (University of Florida), Debbie Keatley (useMyData), Ben Fell (Akrivia), Maksim Belousov (Re:course AI), Richard Dobson (King’s College London)