According to the World Health Organization, 1 out of every 6 deaths can be attributed to one form or another of cancer, making it one of the leading causes of death worldwide. In 2018, 18.1 million people were diagnosed with cancer and 9.6 million sadly lost the battle to this deadly disease. Cancer cases continue to increase globally and although a cure is yet to be found, massive strides both in the cancer research and technology fields have allowed medical professionals to diagnose and treat cases more effectively while also laying the groundwork for drastically reducing the cost of treatment for thousands of patients unable to afford access.
According to Ryan Schoenfield, AI will never replace radiologists and physicians but will contribute significantly towards smaller tasks, such as analysing CT scans and pathology slides. One such AI, Transpara, has already gained FDA approval in the USA and is able to detect cancer in mammograms more effectively than half of radiologists under controlled conditions. During its testing phase, the Transpara System was pitted against 101 radiologists and the system matched the average performance of humans at completing the same task. Cancer cases are increasing worldwide whilst the availability of specialist diagnosticians such as radiologists are in decline. A worldwide shortage of radiologists coupled with these professionals in many cases working in isolation, has led to 25% of breast cancer cases being left undiagnosed. Transpara was developed to combat these staggering statistics by increasing the efficiency and reach of radiologists and improving the accuracy of diagnoses which allows for earlier detection, already showing positive trends towards decreasing the number of breast cancer related fatalities.
Lunit, a medical AI software company founded in 2013, has also developed a variety of AI-driven medical systems in an effort to reduce cancer-related mortality rates. Lunit products are focused on increasing the accuracy of early cancer diagnoses in difficult cases and are able to analyse Chest X-Rays, Mammograms and Tissue Slides, reducing false negatives as well as false positives. Some of their groundbreaking products include the Lunit Insight CXR 1 for lung nodule detection, Lunit Insight MMG for breast cancer detection and Lunit Scope, which analyses blood slides to assist pathologists and researchers in quantifying the amount of tumour infected lymphocytes as well as assist researchers in discovering unusual biomarkers through immune phenotyping.
According to Health IT Analytics, Cleveland Clinic has developed a personalised radiation therapy treatment system that relies on Artificial Intelligence and Machine Learning. The system uses data from medical scans and electronic health records to generate tailored radiation therapy treatment plans and doses according to the patient needs. Radiation therapy is normally delivered to patients without taking into consideration a patient’s individual characteristics and risk factors, but Cleveland’s strategy allows each patient to receive a tailored radiation dose, decreasing negative side effects overall whilst also minimising treatment failures to less than 5%. According to the study Cleveland Clinic published in The Lancet Digital Health, the AI system utilises pre-treatment scan data and extracts tomography scans into a deep-learning model to create image signatures and predict future treatment outcomes. The outcomes are then combined with the patient’s clinical records and risk factors to generate a tailored radiation therapy, which in turn increases the efficiency of the dose whilst mitigating patient side effects.
Cancer treatment and the side effects that normally accompany treatment has also sparked MIT researchers to develop AI-based solutions to treat glioblastoma (brain cancer) patients more effectively. In 2018, MIT Media Lab presented a paper on a self-learning ML model that has the ability to determine an optimal treatment plan to shrink tumour sizes, through lower potency doses. The model is able to learn through reinforced learning techniques, a method normally used in behavioural psychology in which certain behaviours are encouraged or discouraged by means of a reward system.
There is no doubt that Artificial Intelligence has already revolutionised cancer care, but what role will AI play in the future fight against this deadly disease? Whilst some scientists and AI enthusiasts believe we’re at the brink of a cure to cancer, others are a bit more pragmatic. What are your thoughts on the matter? Will this generation be known for curing cancer?