The significant rise of A.I. (Artificial Intelligence) and Machine Learning in recent years has led to profound developments in the world around us, such as the use of ChatGPT to solve problems or the creation of AI-generated art. It can also play a substantial role in medical sciences: in the treatment and diagnosis of many neurological conditions, such as brain tumours. However, the main question that arises is whether A.I. is safe enough to put human lives in the hands of computers.
A new AI technology has been developed at UCL that highlights small tumours and critical structures, such as blood vessels and arteries inside the brain. Trainee surgeons have already started to work with this new technology to learn more precise keyhole brain surgery which will mean more minimally invasive surgeries. This AI system has analysed more than 200 videos of a type of pituitary gland surgery and has reached the level of experience it would take a neurosurgeon 10 years to gain in just 10 months. This means that this AI system has the power to potentially see more operations than any surgeon in their career could ever see, making it robust and highly competent.
Another AI tool developed by Harvard Medical School can allow for faster and more accurate analysis of gliomas (the most common type of diagnosed brain tumour) which reduces the risk of a misdiagnosis by doctors and reduces the time that patients are in the operating room by up to 1 hour. This technology is also more accurate than traditional techniques of analysing the molecular makeup of a specific glioma.
A CT scan analysed by A.I. to show the glioma (shown in pink)
Neurosurgery is an extremely complex field as it requires years of intensive training and a high level of intelligence, decision-making and surgical skills that only a trained surgeon can possess through exposure and experience. Neurosurgeons must be incredibly precise as going even just a few millimetres in the wrong direction can kill or severely injure a patient. Added to this the identification of the boundaries and membranes of tumours is difficult, and surgeons need to make potentially life-threatening decisions on where to stop resecting the tumour. A correct decision can save a life, but the wrong one could lead to the cancer recurring. The AI developed at UCL can help surgeons find the boundary between the tumour membrane and the surrounding brain tissue more accurately with minimal risk to the patient, thereby improving a patient’s survival rate drastically. A quarter of medical errors occurring in neurosurgery are technical errors relating to surgical procedures and could have been prevented if A.I. technology such as this was implemented.
Although A.I. can be an immensely beneficial tool to neurosurgeons, we need to consider the drawbacks of relying on it to make high-stakes decisions such as those in the neurosurgical field. The main problems arise with the data that is being used to train these A.I. technologies. According to the BMC, the majority of neurosurgical cases from 2015-19 were for patients over the age of 45, with the most common types of cases in these age groups being acute subdural hematomas and contusions. This means most of the case data fed to the A.I. will be that of older patients so the A.I. will struggle to make accurate decisions when confronted by a neurosurgical case such as of a 7-year-old. The A.I. might make inaccurate decisions since the brain size of a child is much smaller relative to the brain size of an adult. This can lead to a computed boundary for a tumour that is larger than necessary, due to the A.I. not having been trained sufficiently. This is also known in computer science as algorithmic bias and could lead to the A.I. being biased towards a certain procedure or membrane boundary due to the unrepresentative data it has been trained with. The A.I. would create a treatment plan that may lead to serious injury to the patient if surgeons were to solely follow the decisions of this A.I. technology.
Overall, A.I. is an immensely beneficial tool for surgeons to use in neurosurgical procedures and will become more accurate and developed over time as advancements in technology are made. However, it cannot be relied upon to make complex decisions that can have large consequences. Therefore, the experience and knowledge of a neurosurgeon are key and indispensable, and A.I. should only be used alongside experienced surgeons with their decisions taking higher priority than the A.I.. It is safer to put lives in the hands of highly skilled surgeons rather than just a computer, but it can undoubtedly complement a surgeon's own experience to make surgical procedures more precise and reduce the risk of complications.