Lung cancer can remain silent for a long time, which is part of what makes early detection so difficult. According to GLOBOCAN, it is one of the leading causes of morbidity and mortality worldwide. The five-year survival rate is around 5 to 6 per cent, but the odds can improve to 35 to 36 per cent when the disease is detected at an early stage.
Dr. Manoj Deshmukh, Head of Department, Consultant Radiologist, Lilavati Hospital and Research Center, Mumbai, explains that artificial intelligence is increasingly being used alongside advanced screening to spot suspicious changes earlier and address some of the limitations that come with both imaging tools and human interpretation. Crucially, he says, AI is not here to replace the radiologist. Its value lies in helping the expert make a more accurate and timely assessment.
Why early lung cancer can be difficult to detect
One of the biggest challenges with lung cancer is that it can remain asymptomatic for a considerable period. By the time symptoms such as a persistent cough, blood in the sputum or unexplained weight loss appear, Dr. Deshmukh says the tumour has often already metastasised. This means patients may miss the early and optimal window for treatment.
"The way to address this challenge is through advanced screening tools that can enable earlier detection," he says.
One such screening tool is low-dose CT, or LDCT, which according to the radiologist can have a sensitivity of up to 80 per cent. But higher sensitivity comes with its own challenge: a higher rate of false positives.
A positive screening result can lead to a follow-up scan, additional clinic visits and, in some cases, a biopsy. It can also leave a patient spending months fearing the worst, even if the finding ultimately turns out not to be cancer.
There is another side to the problem. The human eye can occasionally miss subtle patterns or abnormalities on a scan, potentially contributing to a false-negative report. According to Dr. Deshmukh, reducing this burden is one of the areas where AI can help.
How AI analyses lung CT scans
Artificial intelligence models used for this purpose primarily rely on deep-learning algorithms trained using large sets of existing data. Convolutional Neural Networks, or CNNs, are among the main tools involved.
Explaining how they work, Dr. Deshmukh says, "CNNs apply learned filters across the CT volume and build features hierarchically, beginning with edges and textures and progressing to entire lesions."
Once trained on tens of thousands of previously diagnosed scans, these systems can assess lung nodules based on characteristics including their margins, density, shape and changes in volume over time. This process, known as radiomic feature extraction, helps determine the risk of malignancy.
AI-based processing can also involve contrast and window optimisation, helping standardise images for assessment.
According to the doctor, these advances have helped shift the detection horizon from stages III and IV towards stages I and II.
"Overall, the role of AI can range from lung segmentation and nodule detection to reducing false positives, characterising nodules and helping with prognostication," Dr. Deshmukh explains.
Why radiologists still remain essential
Finding an abnormality is only part of the process. An expert radiologist then evaluates and validates the AI-generated findings to arrive at an accurate and timely diagnosis.
This combination, Dr. Deshmukh says, can make diagnosis more reliable while helping reduce unnecessary procedures and false-positive results. Advanced image analysis provides computational support, while the radiologist brings clinical expertise and experience to interpreting what the scan actually means for the patient.
"The collaboration between AI and radiologists can make lung cancer screening faster, more accurate and more effective, ultimately improving the chances of early intervention and successful treatment," he says.
AI complements doctors rather than replacing them
For Dr. Deshmukh, the larger significance of artificial intelligence in lung cancer screening lies in its ability to address some of the shortcomings of traditional approaches.
AI can assist with image interpretation, help reduce false positives, identify subtle abnormalities and support risk stratification. Together, these capabilities can connect earlier identification of the disease with more timely treatment.
But the distinction is important. "Using AI for lung cancer screening does not mean replacing radiologists. It means complementing their work by combining clinical knowledge with computational analysis," Dr. Deshmukh says.
It is this partnership between technology and medical expertise that he believes could significantly change the way lung cancer screening is approached, helping detect the disease earlier while allowing radiologists to make better-informed assessments.
ALSO READ:
Lung cancer in non-smokers: Early recognition and expert prevention advice
