Updated: Apr 4
Organised by the Imperial College School of Medicine Coding Society, the Apply Program Innovate (API) Conference aims to connect students from across the world with the entrepreneurs, researchers, doctors, and engineers that are pioneering innovation in the MedTech industry. Through talks, workshops, and panel events, attendees will be able to learn about new technologies and MedTech solutions which are revolutionising the healthcare industry, inspiring them to get involved in opportunities available to them.
The ongoing COVID-19 pandemic has highlighted the inefficiency of the current healthcare system in the UK, and there is now a need and an expectation for the system to improve.
A huge problem plaguing the NHS is the high cost of healthcare; thus, the goal of innovation in the medical field is to improve accessibility and affordability whilst still maintaining a high standard of care and ensure positive patient outcomes overall. One solution which has the potential to do just this is the automation of healthcare – particularly surgeries.
Now, the debate over the extent to which healthcare should be digitized still stands; however, it should be made clear that the ultimate goal is not to completely automate healthcare - it is to increase the efficiency and quality of the healthcare we receive.
For example, the use of artificial intelligence (AI) in surgeries has the potential to improve patient outcomes in the long-term, which will inevitably result in healthcare becoming more affordable and accessible over time, whether it be through the direct use of AI in real surgical procedures or helping to train and educate current and trainee practitioners in their respective fields.
An example of the latter is the medical device company Medtronic, which is revolutionising the healthcare technology sphere through its development of devices and algorithms, which have applications in modern healthcare. An example of this is the creation of the Touch Surgery App by Digital Surgery (now a part of Medtronic) which is helping to drive the evolution of surgeon training. By using real surgeries to train the AI, it is possible to learn and improve upon existing surgical skills and knowledge. The app offers detailed and accurate descriptions explaining procedures with instructions on how to perform them. There are also various interactive elements so that surgeons and students can learn about procedures and practise them without being directly present in the operating room remotely.
As mentioned previously, the aim of AI in surgery is to constantly improve and optimise the surgical environment, not to completely automate it. Whilst machines may never be able to completely replace humans in healthcare, they can be utilised to measure and optimize pre-existing knowledge, and to ensure all patients receive the same high standard of treatment regardless of external factors. Before developing a product (a solution), it is first crucial to start with the problem which needs solving. A real-world example is the Da Vinci – a powerful robot which is predominantly used to perform minimally invasive surgical procedures, such as hysterectomies and prostate removals. By using real surgeries, machines can be trained to memorize steps taken in surgeries, and perform them with a high accuracy, all while being controlled externally. Since robots can be used for minimally invasive procedures, patients won’t need to heal from large incisions, potentially improving health outcomes for patients post-operation. The effectiveness of AI in healthcare has shown to be promising with one study involving numerous orthopaedic patients, where the study found a five-times reduction in the number of complications resulting from AI-assisted operations compared with surgeons operating alone. The overall improvement in health outcomes can improve efficiency and help drive down overall costs in the long-term because the reduction in complications means less money is spent on resolving further complications and can help reduce a patient’s stay in hospital.
AI can be described (loosely) as the study of computer algorithms giving machines the ability to perform cognitive functions such as problem-solving and decision-making. There are four main sub-categories of AI which are being explored and developed: machine learning, natural language processing (NLP), computer vision and artificial neural networks.
Machine learning gives machines the ability to recognise patterns and use this information to learn and make predictions without explicit programming. Machine learning can therefore be used to predict a known result/outcome or searching for patterns in data. This makes machine learning useful in logistical processes, as they can help identify subtle patterns in large datasets which may be indistinctive to the human eye during manual analysis. Another application of machine learning is reinforcement learning. This is essentially where the algorithm attempts a task and learns from its own successes and failures to improve its efficiency.
NLP is particularly useful in analysing content such as patient medical records and physician documentation. This is achieved through the AI’s ability to learn, recognise, and comprehend human language. In order to achieve accurate analysis using NLP, algorithms must be able to understand more than just simple words – syntax and semantics must also be learnt and integrated into the AI algorithm. If achieved, meaning can be derived from unstructured data (e.g., doctors notes following patient assessment).
While this is constantly improved and requires a lot of time and effort to fully optimize, it can prove useful in making diagnoses more efficient and accurate.
Computer vision is primarily related to the algorithm’s ability to comprehend images and videos and derive meaning from it. This can include object and scene recognition. This has various applications in healthcare, such as in imaging interpretation and acquisition. This can be used in computer-aided diagnoses and image-guided surgery (which is of great interest to MedTech companies). Whilst predictive-video analysis is still under development as it is currently in its early stages, it is clear this has the potential to provide massive amounts of surgical data to predict/prevent adverse events during surgery in real-time, which high-resolution CT imaging may be limited in.
Artificial neural networks are a subfield of machine learning and focus on using biological neural systems to process signals from layers of computational units (in essence, “artificial neurons”) to learn different input-output maps which correspond to human tasks such as image/pattern recognition and data analysis. By using multiple layers of neural networks and signals, complex human behaviours can be understood in greater detail than algorithms using one or two layers of data. This offers exciting potential applications in healthcare, due to their high sensitivity and specificity in predicting responses following interventions.
By combining various aspects of AI, we can develop applications and improve existing interventions for a plethora of healthcare-related issues.
While the true potential for AI is difficult to predict at the moment, due to its infancy in healthcare, it is clear that synergy across fields can massively improve the efficiency and accessibility of healthcare systems globally, whilst maintaining (or potentially even improving) health outcomes for the population. And while the prospects for AI in healthcare are promising, it is clear that there are various limitations of it as well. Since the training of AI algorithms depends upon existing data, the type and accuracy of this data must be questioned before it is fed into algorithms. This can have a disproportionate impact on those of currently under-represented backgrounds, such as those of BAME ethnicities and women. As well as this, AI is currently unable to differentiate between correlation and causation. AI cannot determine causal relationships from data at a level of clinical significance and requires human interpretation to derive. The inability of AI to apply appropriate clinical context to data limits its use in data analysis and interpretation.
Overall, there is great financial incentive to apply AI technologies to healthcare due to their potential in the medical field. For example, it would be possible to collate and pool knowledge globally regarding procedures and develop a universal approach and standard to healthcare. This can improve patient outcomes on a global scale for all patients. This can help reduce inequalities in healthcare and treatment. However, they cannot completely remove the need for medical professionals.
The need for human interpretation and human development means that healthcare cannot be completely automated; nor should automation be the goal of healthcare.
Ultimately, it is the professional providing clinical information to patients, and only by establishing a doctor-patient relationship can we treat the patient, not the disease.