ArtificIal Intelligence in Orthopaedic Trauma

Why?

Nobel prize winner, psychologist Daniel Kahneman, reveals important human biases in our day-to-day decision making.

Similarly, in Orthopaedic Trauma surgical decision making may suffer from surgeon biases.

For example, the prediction of adverse surgical events -risk stratification- is very challenging. One could argue that when one unique patient consults us for their chance for the bone to heal, or the probability that this healing process will be complicated by an infection, our personalized advice for one’s patient is an educated guess at best. 

Also, fracture recognition and classification may be biased, as different surgeons see different things. Or to popularize: “Surgeons agree mostly with themselves (i.e. satisfactory intra-observer reliability), but not so much with each other (i.e. poor or fair inter-observer reliability)”. Inter-surgeon agreement does improve with more advanced imaging modalities, but surgeons persist to see different things, even when looking at the same 3D hand held models of a complex fracture. 

Artificial Intelligence may augment clinical decision making by overcoming certain surgeon biases.

What?

Artificial Intelligence (AI) is believed to change the scope of Medicine, as the introduction of smartphones changed our day-to-day life. Indeed, deep-learning algorithms show promising results as valuable diagnostic tools to assist clinicians in many respective specialties.  

In Orthopaedic Trauma, studies on applications of AI technology are scarce but exciting: in the field of Predictive Modelling, Machine Learning driven probability calculators are promising tools for risk stratification:  accurate personalised prediction of the chance to develop an infection after a tibial shaft fracture.

In the field of Computer Vision, Convolutional Neural Networks (CNN) perform on a human level in recognising fractures on plain radiographs taken in the Emergency Department, and CNNs are even able to classify and characterise complex fractures reliably.

How?

The Machine Learning Consortium aims to: 

1) connect curious surgeons and expert computer scientists in a collaborative effort to ask the right questions, as well as to 

2) connect international data sets to answer these clinically relevant questions in order to a) train, b) test, c) externally validate; and d) prospectively evaluate CNNs and ML probability calculators in an international collaborative effort in the field of Orthopaedic Trauma.