Dr Sumeet Hindocha
Clinical Oncology Specialist Registrar & PhD Candidate in Artificial Intelligence, The Royal Marsden Hospital & Imperial College London
It’s estimated that 50% of people will experience cancer at some point in their lives. New AI technology could play a pivotal role in identifying at risk patients early on and improving cancer diagnosis.
Earlier diagnosis is linked to better outcomes and the NHS has set out an ambitious priority for 75% of patients with suspected cancer to have a diagnosis within 28 days by 2028.
Workforce pressures
COVID-19 has placed huge pressures on the health service and coupled with longstanding workforce shortages in primary care, oncology and diagnostic specialities, such as pathology and radiology, will make realising this goal even more challenging.
The Royal College of Radiologists estimate a 44% shortfall in the number of radiologists and 29% shortfall in the number of clinical oncologists by 2025. Whilst important in the longer term, owing to the length of specialist training required, increasing recruitment will not be enough to solve immediate workforce pressures.
Diagnosing cancer can be tricky as early symptoms may be vague and overlap with less serious conditions.
Enhancing cancer diagnosis with AI
Innovations in artificial intelligence (AI) present a promising solution. Research continues to demonstrate numerous examples of AI being used to improve cancer diagnosis. Diagnosing cancer can be tricky as early symptoms may be vague and overlap with less serious conditions. GPs are often the first port of call for patients with worrying symptoms and here, AI tools are being piloted to help identify which cancer a patient may be at risk of and suggest the next best step such as blood tests, a scan or urgent referral.
Improved patient outcomes
Patients referred for a scan or biopsy require radiologists or pathologists to interpret and report the results. Shortages in these specialities and significant backlogs are leading to delays in diagnosis. AI is being employed to triage investigations to determine which require urgent specialist review so that the large volume of normal tests are not unnecessarily escalated.
In breast cancer, AI tools to detect abnormalities on mammograms have shown performance comparable to radiologists. Similarly, tools that can determine which abnormalities on chest CT scans are likely to become lung cancer, or which can automatically detect prostate cancer from a biopsy are entering the clinic.
Of course, these exciting advances should not be met with pure, unguarded enthusiasm. They require robust validation and regulation to ensure safe and effective use. However, by facilitating data analysis at a level far beyond the limit of human capability, AI has the potential to dramatically enhance cancer diagnosis and lead to better outcomes for patients.
