Practical benefits of AI – replacing doctors?

Our clinical experts are happy to shed some light into this topic by discussing how AI approaches actively support clinicians to elevate the healthcare industry to higher standards. At ImageBiopsy Lab this is no different. Our mission is to support all MSK-related imaging in order to facilitate orthopedists and radiologists in their day-to-day work.

Authors: Christoph Salzlechner and Allan Hummer

On December 17, 1903, Orvill Wright performed the very first actively powered aeroplane flight, starting a whole new era of travelling [Encyclopaedia Britannica]. Over a century later, commercial flight has ascended to a quite different level, transporting passengers and goods rapidly around the globe. With a whooping 4.5 billion passengers transported in 2019, the aviation industry represents a central aspect of modern civilisation [International Civil Aviation Organization]. This has been possible thanks to technological advances, especially due to the integration of the auto-pilot technology. Unlike the initial implication, the auto-pilot system is not operating the aeroplane independently, but rather assisting and supporting the pilots by computing complex and time-consuming tasks [Federal Flight Administration]. Undoubtedly, this has improved flights in safety, efficiency and economical value, making the auto-pilot a widely accepted piece of technology.

This development, which enabled the rise of aviation holds an analogy that is in parallel to how radiographic imaging is developing. 

Starting with the discovery of X-rays on November 8, 1895 by Wilhem Röntgen, modern radiography represents a fundamental feature of medicine and medical research [The Nobel Prize Foundation]. Since 1895, imaging processes have been in the focus of development, which resulted in increased image quality and the advancement of rapid imaging techniques such as computed tomography. By using other physical phenomena, imaging modalities such as magnetic resonance imaging and ultrasound imaging became possible [Bercovich et al].

Nowadays, medical imaging represents non-invasive and pain-free approaches, which enable accurate diagnosis, disease prevention and selection of proper treatment. The WHO and PAHO estimated that about 3.6 billion diagnostic X-ray examinations were performed in the year 2012 worldwide [WHO/PAHO]. In order to tackle the surplus of information available, the digital revolution, starting a few decades ago, has also affected the medical sector. More recently, artificial intelligence (AI) has come into focus in academic and industrial research. Machine-learning approaches, like Computer-Aided Diagnosis, have found great potential in assisting with the detection of potentially fatal brain hemorrhages or lung nodules [Chartrand et al, Erickson et al]. These have proven beneficial for clinical diagnosis, as through the evolution of imaging techniques, the amount of data has become time-consuming to process for clinical users. Just like the auto-pilot in an aeroplane, this technology is not meant or even capable of  replacing a clinical expert. This is due to the fact that the AI’s algorithms require precise training and are limited to address a quite specific task. However, there are various benefits when integrating AI into the clinical workflow. 

One of the main medical areas AI is currently employed in is orthopedics. Knee osteoarthritis (OA) presents itself as a degenerative disease that affects an estimated 6% of adults over 30 years of age and 13% of adults over 60 years of age [Felson et al]. The diagnosis of knee osteoarthritis is most commonly performed by observation of plain radiographs and assessment of different features of the tibiofemoral and patellofemoral joint via the Kellgren & Lawrence scheme (KL). This gold standard KL grading system does have inconsistencies in the description of radiographic features of OA and the prominence of certain factors in various stages does not necessarily reflect the patients’ OA development. This makes it difficult for clinicians to assess and track disease progression, decision on ideal treatment options and further results in high inter-reader variability [Spector et al, Schipof et al, Yunus et al]. Here, AI has the potential to standardise analysis and objectively present the clinical user with reproducible values, leading to increased inter-reader agreement [Nehrer et al]. This is not only resulting in swifter analysis time of radiographs and more precise treatment decisions, but also presents the potential to predict disease progression [Paixoa et al].

Besides direct clinical support via image analysis, AI is further employed in mobile health technologies (eHealth) as well as value-based payment models [Haeberle et al]. Forbes Magazine has reported that by 2026, the use of AI in healthcare applications will result in annual savings of $150 billion [Forbes Magazine].

Like in the aviation industry, novel technologies like AI approaches actively support clinicians to elevate the healthcare industry to higher standards.