Among the cancer-related deaths worldwide, lung cancer is the most frequent cause. Lung cancer is diagnosed in 1.8 million people each year, and 1.6 million people die per year due to the disease. According to data provided by Cancer Research UK, updated in 2020, the net five-year or more survival rate is 13.8 %. The tertiary prevention of lung cancer has been influenced by studies on its relative risk factors and the epidemiological characteristics of lung cancer. 

Seventy-five per cent of lung cancer cases are detected only in its advanced stages with nodal spread and metastatic disease due to the minimal or absence of symptoms in the initial stages of the disease. This can influence the survival rate among patients diagnosed with lung cancer. The National Lung Screening Trial (NLST) findings implicated that screening for lung cancer in high-risk individuals using low dose computed tomography can reduce lung cancer mortality by 20 %. 

AI for lung cancer 

A new UK study has revealed the potential of using AI to help doctors diagnose lung cancer earlier. LIBRA, led by researchers from the Royal Marsden NHS Foundation Trust, the Institute of Cancer Research, London, and Imperial College London, used data from the CT scans of nearly 500 patients with large lung nodules- abnormal growths- to develop an AI model. 

The study was supported by Royal Marsden Cancer Charity, the National Institute for Health and Care Research, RM Partners and Cancer Research UK. To assess the model's effectiveness, the team used a measure called the area under the curve (AUC). An AUC of one would indicate a perfect model, while 0.5 would be expected in the model that was randomly guessing. 

The results, published in the Lancet's biomedicine, indicate that the model could identify each lung nodule's risk of cancer with an AUC of 0.87. In addition, the performance improved on the Brock and Herder score- two tests currently used in the clinic. 

The model can also aid doctors in making quick decisions about patients with abnormal growths that are currently deemed medium risk by current tests. When combined with Herder, the model was able to identify high-risk patients in this group and would have suggested early invention for 82% of the nodules that went on to be diagnosed as cancerous.  

Earlier models 

The UK model for lung cancer detection is not the first of its kind. Earlier this year, MIT developed a stand before a CT scanner at MGH, where they generated part of the validation data. The model calculated the patient's chance of developing lung cancer.  

Similarly, as part of the e strategic collaboration, AstraZeneca's Emerging Markets Health Innovation Hubs partnered with Qure.ai to explore the application of deep learning algorithms to identify patients with suspicious radiographic lung abnormalities and support their referral to arrive at a firm diagnosis.  

Artificial Intelligence consistently proves to be a promising advancement. Almost all the studies concluded that incorporating AI into radiology will promote improved patient care by earlier and more accurate detection of the disease, and thereby a good prognosis.  

Due to the better discrimination and evaluation of a greater proportion of lung nodules, cancers being missed can be reduced. In addition, the development of various artificial intelligence algorithms benefitted thoracic imaging for various conditions. 

Limitations 

A report presented at the National Library of Medicine identified some of the key limitations of some of the current AI models. According to the report, a smaller sample size of studies and non-validated models are the main constraints limiting the research in this field. In addition, constraints on computation and incomparable studies are also limiting in nature.  

Statistically significant results are not produced in some studies as the sample size is small. A systematically validated and confirmed model of convolutional neural network or machine learning is still not available for routine clinical use.  

However, Dr Benjamin Hunter, clinical oncology registrar at the Royal Marsden NHS Foundation Trust and clinical research fellow at Imperial College London, who is funded by Cancer Research UK, believes that in the future, their AI model will improve early detection and potentially make cancer treatment more successful by highlighting high-risk patients and fast-tracking them to earlier intervention. Going ahead, they plan to test the technology on patients with large lung nodules in the clinic to see if it can accurately predict their risk of lung cancer.