Lung nodules

What are lung nodules, and how common are they?

Lung nodules or pulmonary nodules are small spots in the lung that are less than 3 cm in size and are often found during imaging scans. Each year, over 10 million lung nodules are detected worldwide, with more than 2 million found in the US alone.¹ While most are harmless (benign nodules), some are an early sign of lung cancer (malignant or cancerous nodules), making an assessment of them critical for early detection of lung cancer.

Nodules vary widely in shape, density, and size. Smooth, round nodules are often benign, caused by infections, inflammation, or scar tissue. In contrast, spiculated nodules, which have irregular, star-like edges extending into the surrounding lung tissue, are more likely to be cancerous. In terms of density, nodules may be solid, part-solid (with both solid and hazy components), or pure ground-glass nodules (GGNs). GGNs appear hazy and can be either benign or early-stage cancer. Size also matters; nodules under 6 mm are usually benign, while those over 8 mm have a higher risk of being cancerous. If a cancerous nodule grows unchecked, it can develop into a lung mass or tumor.³

What causes lung nodules?

Lung nodules form due to various causes, including infections, inflammation, and scarring from conditions like tuberculosis. Environmental exposures, such as air pollution and occupational hazards, can also contribute. In around 3-5% of cases, nodules result from early-stage lung cancer.² Distinguishing between benign and cancerous nodules is a critical step in lung cancer detection, as failing to do so can delay diagnosis. Studies estimate that each year, around 300,000 patients with lung nodules that are cancerous are initially missed, underscoring the challenge of identifying high-risk nodules.³

Are all lung nodules dangerous?

No, the vast majority of lung nodules – more than 95% – are benign and caused by infections, inflammation, or old lung injuries. However, the remaining 5% can be cancerous or develop into lung cancer over time.^1,2 Cancerous nodules tend to grow and change over months, whereas benign ones usually stay the same size. If left undetected, a cancerous nodule can grow into a lung mass or spread beyond the lung. The challenge for doctors is to accurately determine which nodules need further testing while avoiding unnecessary procedures for benign cases.

How are lung nodules detected and diagnosed?

Lung nodules are most commonly detected through chest X-rays or CT scans, often found incidentally when patients are scanned for other reasons. However, X-rays often lack the resolution needed to detect small nodules or accurately characterize their features, which can lead to missed diagnoses. CT scans provide much more detailed images, allowing doctors to assess a nodule’s size, shape, density, and growth over time.^4-6

In the US, once a nodule is detected, a radiologist assess its morphology and documents their findings in a radiology report, which is sent to the referring clinician. If the nodule requires further evaluation, the patient may be referred to a nodule clinic. In the nodule clinic, a nurse navigator (a registered nurse who serves as a care coordinator) reviews the patient’s nodule(s), coordinates appointments, and facilitates communication among a multidisciplinary team that may include a pulmonologist, thoracic surgeon, and a thoracic oncologist. This team assess the nodule further to estimate the risk of cancer. Nodules that are larger than 8 mm, have irregular borders, or show signs of growth over time are more likely to be cancerous and require additional testing.⁵

In other countries like the UK, the management of lung nodules is also typically overseen by a multidisciplinary team, comprising respiratory physicians, radiologists, and specialist nurses. Upon referral, specialist nurses assess CT images to determine the nature of the nodules and decide on the necessity of follow-up scans or referrals to lung cancer specialists. PET scans, tissue biopsies, genetic testing, and advanced AI-assisted imaging tools are then used to diagnose and further characterize any cancerous lung nodules.⁶

Why not just assess all these nodules then?

With more than 10 million lung nodules found each year, assessing each nodule individually and manually would overwhelm health systems. Therefore, determining whether nodules are cancer or not cancer is a big challenge, with many cancer patients failing to be ruled in and diagnosed with cancer in a timely fashion, and many other patients failing to be ruled out, and thus undergoing unnecessary invasive procedures (such as biopsies) to diagnose whether their nodules are cancerous.⁷ While it’s very obvious to doctors that some nodules are not cancerous, in many cases, it is very difficult, even for experts, to tell whether they are benign or cancerous without further investigation. There are a lot of these indeterminate pulmonary nodules (IPNs), but not enough nodule experts to assess them. And they make the task facing health systems even harder.

Can AI lung nodule detection help?

On their own, AI‑based computer‑aided detection (CADe) software risk overburdening clinicians and health systems with too many indeterminate pulmonary nodules. While they can highlight suspicious nodules that a radiologist might otherwise miss, there are numerous studies pointing to their high false positive rate.^8-10 Some recent studies show false positive rates as high as 39% and 40%, respectively.^9,10 A high false positive rate means that there are a high number of lung nodules detected that are benign. Every one of these "false alarms" has to be opened, inspected, and dismissed by a human reader. Moreover, they can increase the risk of unnecessary biopsies and procedures for patients.⁸ So, using CADe as a first‑pass screener can actually increase workload rather than reduce it. This is why CADe software is often best paired with AI-based computer-aided diagnosis (CADx) software. CADx software helps assess the risk of malignancy in nodules, stratifying and prioritizing patients based on their risk profile.¹¹ In short, AI detection can help clinicians catch more cancerous nodules, but without AI risk assessment, it cannot solve the triage problem created by millions of indeterminate nodules.