Although AI is often thought of as synonymous with robotic virtual assistants such as Siri or Alexa, or developing personalised suggestions for Google searches, adverts or programmes on platforms such as Netflix, and is often regarded in many cases with suspicion or even outright fear by the general public, there have been some developments in its application which should be warmly welcomed by the population at large: The use of AI in the diagnosis of medical conditions, especially the most serious ones such as cancer and degenerative conditions such as Alzheimer’s.
At Polestar we have noted a number of developments suggesting greater alignment between the healthcare and technology industries, including the application of AI and big-data-focussed solutions. Generally, the healthcare’s data-heavy nature makes it an ideal candidate for the application of AI across a range of disciplines, from diagnosis and pathology to drug discovery and epidemiology.
Early detection is key in the fight against cancer and in too many cases individuals are put off going to see their doctor due to the inconvenience or the difficulty getting an appointment (and then not attending their follow-up appointments), or symptoms are incorrectly attributed to another condition. Early or ‘pre-cancer’ symptoms can be difficult to recognise and current methods rely on human analysis of scanned images, limited information from hereditary testing, or combining multiple diagnostic techniques which can be invasive and costly.
AI could drastically transform the assessment of cancer to a quick, simple, cheap and accessible option that could mean many more cancers are caught in their earliest stages and effective treatments started before there is any spread of the disease.
In the past, AI solutions have struggled to penetrate the healthcare field as neural networks require large amounts of data in order to refines themselves and getting enough reliable data is major barrier; medical data is often heavily protected and is not always collected consistently enough. The general rule that a sample size needs to be 10x greater than the features searched for has meant that it application in medicine has been near impossible. Nonetheless, some companies are overcoming these challenges in unique ways, including training algorithms in reverse and using old-fashioned AI to explore completely new fields of medical diagnosis.
There have been several promising success stories so far:
Lunit – Image detection
Based in in South Korea, Lunit has trained their INSIGHT algorithm on chest x-rays and mammography images to detect lung and breast cancer. In 2016, they trumped Microsoft and IBM to win the Tumor Proliferation Assessment Challenge, boasting a 97% detection rate for lung and breast cancer. The CEO note that ‘it is difficult for doctors to find small nodules hidden behind ribs or organs in chest x-rays’, but their algorithm searches extensively for cancer patterns in order to drastically reduce the chance of a false negative or missed case of cancer. The team used CT scans that were biopsy-proven, therefore eliminating human bias.
Quantgene – Targeting cancer molecules
Quantgene is part of a new field of DNA-centric approaches to cancer detection, seeking to combine the two fields of cancer screening and genomics. Their algorithm analyses cell-free DNA to determine the genetic disposition to cancer, whether there is any cancer currently present and in what part of the body. Cell-free DNA (cfDNA) is expelled into the bloodstream from cells that have died or been killed, and they make up around 10,000 cells in a typical blood sample, which might contain ten million blood cells. Quantgene’s algorithm is designed to analyses every individual copy of cfDNA in a sample to determine if any single copy resulted from a tumour cell.
Lancor – Tumour Trace
London-based Lancor Scientific has designed a diagnostic device that combines AI with Opto-magnetic Imaging Spectroscopy (OMIS) technology, which is based on electromagnetism. When light is shone onto human tissue, the magnetic component of the reflected light can determine whether there is any malignancy. Current equipment required for this type of testing – magnetic and atomic force microscopes – are both huge and expensive, limiting their penetration. Lancor’s Tumour Trace uses the same principle as these large machines, but employs light rather than radiation, with the device weighing c.5kg and a single test costing around £10. During trials at Southend Hospital in 2018, Tumour Trace detected cervical cancer with more than 90 per cent accuracy. By using AI to remove biological noise from the signals, it is believed this can be improved to 97 per cent, and that the device could eventually be used to screen for all cancers.
The application of AI is not just limited to oncology, with DeepMind Health working with Moorfields Eye Hospital, training software to diagnose a range of ocular conditions from digitised retinal scans. That work resulted in an AI system able to recommend the correct referral decision for over 50 eye diseases with 94 per cent accuracy, matching the performance of top medical experts.
It has been stressed that AI technology won’t replace medical professionals, with computer algorithms designed to assist humans with more difficult diagnoses. In line with the UK government’s Industrial Strategy, five new AI research centres were announced in November last year, to be located in Leeds, Oxford, Coventry, Glasgow and London. Backed by a £50m investment and due to open in 2019, these centres will focus on the rapidly advancing area of image analysis.
At Polestar, we have experience in advising technology companies which are assisting the healthcare industry in both collecting and analysing the vast volumes of data it generates. The application of AI in healthcare is likely to be one of the most positive application of this technology channel. Wide-scale implementation of AI could lead to a proactive healthcare system which responds to diseases preemptively rather than focusing on treating already ill people.
“If applications like Lunit and Quantgene can make themselves known – and more importantly understood – by the healthcare community at large, then AI will be a powerful weapon in the fight against cancer.”
