✕
☰

Beating cancer – the AI breakthrough

by

Prashant Warier, co-founder & CEO, Qure.ai, writes about how AI has become key when tackling cancer. 

Challenges facing radiology

A string of reports from the Parliamentary and Health Service Ombudsman service, the Royal College of Radiologists (RCR) and Professor Sir Mike Richards’ Independent Review of Diagnostic Services for NHS England have indicated major concerns for radiology to address due to the need built up over the pandemic. It’s clear from these report that there are significant obstacles to overcome, notably dealing with the backlog in demand for scans and optimizing workflows for doctors and radiologists. 

According to the NHS data there were 34 million imaging tests reported in England in the 12 months from April 2020 to March 2021 – this contrasts with 44.9 million imaging tests reported in England in the year to March 2019. The Taskforce for Lung Health estimated that 20% of lung disease patients waited for more than a year to receive a prognosis. NHS England data suggested that almost 400,000 patients were waiting for more than a year to start cancer treatment – the highest figure since December 2007.

The other major challenge facing radiology was outlined by Rob Behrens, chair of the Parliamentary and Health Service Ombudsman service and highlighted in his report: “X-Ray and scan results are key to diagnosis and treatment for many people. Yet the failings outlined in this report show that, without a concerted effort to improve imaging, patient safety continues to be at risk.”

The core issue is one of disjointed care cascades, which boils down to interoperability in healthcare. During the patient's treatment, they may see a GP, radiologist, chest specialist, and an oncologist. Each will add to the patient's analysis and diagnosis, resulting in a paper trail – if this data is not collected in one location, the diagnosis and, eventually, treatment, will be delayed.

The findings were welcomed by the RCR because they demonstrated the results of a study it had undertaken on attitudes among radiologists in the United Kingdom. 

In his report, Professor Sir Mike Richards, the former NHS Cancer Tsar, laid out the scope of the issues that needed to be addressed. He predicted that the NHS' CT scanning capacity will need to increase in the next five years. 

Solving backlog, reporting, and healthcare interoperability issues is difficult, and no single solution can handle all of radiology's operational, technological, and personnel challenges. However, we at Qure.ai believe that artificial intelligence (AI) has the potential to reduce inefficiencies and optimise workflows, resulting in increased productivity and an additional layer of confidence for physicians.

Using AI for lung cancer screening

Lung cancer is the most common type of cancer across the globe. Its aggressive nature and absence of early, visible signs often leads to diagnostic and treatment delays and poor outcomes. Although common risk factors like smoking are significant contributors to lung cancer, aspects such as genetic susceptibility, poor diet, occupational exposure, and air pollution can make it challenging to determine the likelihood of lung cancer in patients. 

Early diagnosis of the disease greatly increases the odds of survival and allows for the adoption of less invasive medication or intervention. Systematic population-based screening aids in the diagnosis of abnormal pulmonary results, the most effective use of confirmatory testing, and cancer detection in the early stages. Early detection benefits individuals, their families, and society. 

However, early identification of lung cancer is difficult because: 

AI can address the challenges facing radiologists by automating the process to detect malignant lesions. It can help clinicians with the issues of healthcare interoperability, such as misdiagnosis, analysis, reporting, detection of co-findings and reducing time constraints. 

A typical patient journey

Regarding lung cancer, there are protocols in place requiring regular screenings for at-risk patients. In one possible scenario, a GP might spot an area of concern in an X-Ray but be undecided about referring the patient. If the AI is involved in the workflow, it can flag any concerns when looking for different findings including lung nodules.  

Once the patient is hospitalised the AI can help to follow up on the scans and help to track lesions over time. It can spot suspect nodules using algorithms based on parameters recommended by radiology and cancer societies. It can generate a score to assess the risk of a nodule being cancerous, which can be done for the different types of lung cancer – non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC). 

The technology has also been designed to be explainable and modifiable AI. That means the decision-making process is intelligible, not hidden away in a metaphorical black box that cannot be interpreted. For example, we have built our software so that it can illustrate findings in the same way that teachers in medical school articulate the problem. The AI points out what is observed with annotations, which is then integrated into the reporting workflow. The decision to use the observation or edit it lies with the radiologist or clinician who is responsible for taking the next steps in the pathway.

A clearer picture

Everyone concerned with lung cancer treatment very much understands the importance of early detection and diagnosis. The most important benefit of an AI solution is speed of diagnosis, as the cost of treating lung cancer cases alters depending on how early the cancer is spotted and treatment begins. Instant results let referrals be fast-tracked for further investigations, allowing physicians to optimise their time and relieving patients of anxiety if scans are normal. Not only does this help keep the costs of screening to a minimum, but also reduces the burden of direct and indirect costs on the individual and the economy. 

Back to topbutton