Radiological Expertise

Summary:

• Research into aspects of radiological expertise include cognitive and perceptual studies
• Cognitive studies have identified the same kinds effects in radiology as found in other medical decision making tasks
• One study also suggests that verbal prompts can impede the processing of visual information
• Perceptual studies reveal that the acquisition of radiological expertise is in part a form of perceptual learning in which low-level detectors in the visual system become finely tuned for particular features
• Different radiologists will look at different regions of the image. If we use computers to analyse the regions that attract a radiologist's attention, we can compute features which can be used to train a neural net to predict how he or she will classify a given region.

 [Presentation (1.4 MB) ]

Medical Logics

Summary:

Review of Logic based reasoning in Medical AI systems:

• Reasoning about what is the case - beliefs (e.g., diagnosis)
• Reasoning about what one ought to do - actions (e.g., medical treatment planning)

Review of state of the art in application of logic based models of argumentation (beliefs and actions) in medical AI; how these models can:

• facilitate decision making
• support collaborative decision making
• enhance communication between medical professionals and patients, as well as inform education of medical professionals

 [Presentation (194 KB) ]

Medical Planning

Summary:

Much of clinical practice can be understood as either creating, modifying or interpreting plans (treatment plans and protocols, diagnosis plans, "integrated care pathways" etc).

Despite the central role of plans in medical practice and medical knowledge, and the considerable literature on computerised clinical plans and protocols, very little work has been done on how clinicians plan or interpret plans.

Planning is known to be very demanding in terms of cognitive resources. Like other experts, clinicians acquire large numbers of simple, stereotyped plans during their training and practice, which they can adapt to different situations encountered in practice to reduce the need to generate plans from scratch.

These plans or schemas are likely to be hierarchically structured and centered around goals and sub-goals, making them easy to adapt to specific situations.

Full-blown planning from scratch is likely to be too difficult to carry out routinely without assistance. It involves at least the following types of high cognitive load:

1. The developing plan must be held in memory (human short-term memory is well known to be severely limited).
2. At each step in the plan, the various available options for action must be identified.
3. The pros and cons of each option must be identified, and one chosen.
4. Constraints on & dependencies between planned actions must be tracked as the plan develops. This becomes exponentially more difficult as the plan becomes more complex.
5. The overall effectiveness of the developing plan with respect to its goals (e.g. to reduce risk or improve outcome) must be tracked.

REACT ("Risks, Events, Actions and their Consequences over Time") is a software system developed at CRUK to support each of these cognitive loads during planning. It has proven effective in situations where clinicians must develop care plans jointly in collaboration with patients.

 [Presentation (627 KB) ]

Medical Thinking: Towards A Unified View

Conclusions:

• Humans are fallible (individually and organisationally)
• What do we know that can help?
• Answer depends on whether you are a scientist, a theoretician or an engineer; all these viewpoints have a contribution
• We have reviewed medical thinking as: knowledge representation and use, the application of “rational”logical methods, decision-making and planning as tasks that can be rigorously engineered
• PROforma unifies these perspectives in a formal but natural and versatile model
• Evidence suggests that technologies based on PROforma are useful andprofessionally acceptable

 [Presentation (1.4 MB) ]

Medical Thinking or Clinical Action

The context:

• The NHS costs £80Bn per annum; there are severe workforce pressures
• For some tests and therapies, we know enough about what helps patients to recommend that their use should be reduced or increased
• Despite this evidence, there is much geographical variation in clinical practice and patient outcomes

The issue:

• How can we narrow the gap between what clinicians know and what they do ?

 [Presentation (1.0 MB) ]

From Medical Error To Safety Architectures: Implementing Patient Safety Programmes

Summary:

The England and Wales NHS National Patient Safety Agency (NPSA) was created in 2001 and incorporates the world's first national patient safety reporting system. The goal of the NPSA is "to improve safety of patients by promoting a culture of reporting and learning from patient safety incidents affecting patients receiving National Health Service funded care" Tasks include to:

• "Collect and analyse information on adverse events from local NHS organisations, NHS staff and patients and carers;"
• "Assimilate other safety-related information from a variety of existing reporting systems and other sources in this country and abroad;"
• "Learn lessons and ensure that they are fed back into practice, service organisations and delivery;"
• "Where risks are identified, produce solutions to prevent harm, specify national goals and establish mechanisms to track progress."