AI is increasingly being used in various fields within healthcare, offering innovative solutions that improve patient outcomes and streamline clinical workflows. These applications include assisting in disease detection, enhancing the efficiency of medical chart preparation and automating routine tasks through advanced generative AI chatbots.
Examples you may know already
Medical Image analysis
AI is increasingly used to identify areas of interest in medical images, such as detecting tumors in radiology scans. By analyzing vast amounts of imaging data, these AI systems can highlight suspicious areas for further examination by healthcare professionals, improving early detection and diagnosis.
Virtual Health Assistants
AI-powered virtual assistants are now helping patients manage their health by providing reminders for medication, tracking symptoms and offering advice based on individual health data. These assistants can enhance patient engagement and ensure timely intervention when necessary.
Examples of Research at the University of Southampton
Cancer Treatment Pathway Prediction
A machine learning-based clinical assistive decision tool has been developed that assist multidisciplinary teams (MDTs) in making treatment pathway decisions for curative oesophageal cancer patients. The tool provides data-driven recommendations, standardises decision-making and improves efficiency within the MDT workflow, running parallel to MDT meetings to offer objective patient case assessments and prioritise caseloads. Trained on a dataset of 399 patients, the machine learning model predicts treatment pathways using clinicopathological variables, ultimately aiming to reduce variability in decisions and enhance health equality.
Patient Deterioration Risk Monitoring
The Computer-Assisted Risk of Deterioration Score (CARDS) is an advanced algorithm tailored for critically ill patients on hospital wards, providing real-time assessments of clinical acuity using vital signs and laboratory tests. Designed to support healthcare professionals in identifying patients at risk of deterioration, CARDS aims to improve outcomes by facilitating timely ICU transfers and mitigating adverse events like cardiac arrests and septic shocks. Developed and validated on a cohort of over 6,000 patients at the Ronald Reagan UCLA medical centre, CARDS utilises multitask Gaussian Processes and transfer learning techniques to enhance predictive accuracy and adaptability across diverse patient profiles and conditions.
The open research visions relating to AI within healthcare are as follows:
- Equitable Access to AI-Driven Healthcare Solutions: Ensure that AI technologies are accessible and beneficial to diverse populations, addressing disparities in healthcare availability and quality, particularly in under-resourced regions.
- Transparency and Collaboration in AI Development: Promote open collaboration among researchers, clinicians and policymakers, with an emphasis on transparency in AI algorithms and data usage to foster trust, reproducibility and continuous improvement in healthcare applications.
- Ethical and Safe Integration of AI: Develop and implement AI solutions that prioritize patient privacy, data security, and ethical considerations, ensuring that AI systems enhance clinical decision-making without compromising patient care or safety.
Educational program pathway at the University of Southampton
The biomedical engineering course at Southampton is closely linked to the real-world research being carried out at the university. Students get the chance to engage in a large number of exciting AI-related projects, and learn the necessary skills through modules such as:
- ‘Machine Learning Technologies‘ and ‘Natural Language Processing‘, in which students learn how to extract information from both complex health-related data sets and written text (such as written medical records)
- ‘Bioinformatics and Systems Biology‘, in which students learn about the complex systems that make up the human body, as well as hands-on skills to process data about these systems
- ‘Knowledge graphs for AI systems‘, in which Students learn how to handle data arising from a complex network of linked data, such as that often found on the Web.