Researchers have used a deep-learning algorithm to detect lung cancer accurately from computed tomography scans. The results of the study indicate that artificial intelligence can outperform human evaluation of these scans.

doctor looking at lung scans on computer screenShare on Pinterest
New research suggests that a computer algorithm may be better than radiologists at detecting lung cancer.

Lung cancer causes almost 160,000 deaths in the United States, according to the most recent estimates. The condition is the leading cause of cancer-related death in the U.S., and early detection is crucial for both stopping the spread of tumors and improving patient outcomes.

As an alternative to chest X-rays, healthcare professionals have recently been using computed tomography (CT) scans to screen for lung cancer.

In fact, some scientists argue that CT scans are superior to X-rays for lung cancer detection, and research has shown that low-dose CT (LDCT) in particular has reduced lung cancer deaths by 20%.

However, a high rate of false positives and false negatives still riddles the LDCT procedure. These errors typically delay the diagnosis of lung cancer until the disease has reached an advanced stage when it becomes too difficult to treat.

New research may safeguard against these errors. A group of scientists has used artificial intelligence (AI) techniques to detect lung tumors in LDCT scans.

Daniel Tse, from the Google Health Research group in Mountain View, CA, is the corresponding author of the study, the findings of which appear in the journal Nature Medicine.

Tse and colleagues applied a form of AI called deep learning to 42,290 LDCT scans, which they accessed from the Northwestern Electronic Data Warehouse and other data sources belonging to the Northwestern Medicine hospitals in Chicago, IL.

The deep-learning algorithm enables computers to learn by example. In this case, the researchers trained the system using a primary LDCT scan together with an earlier LDCT scan, if it was available.

Prior LDCT scans are useful because they can reveal an abnormal growth rate of lung nodules, thus indicating malignancy.

In the current study, the AI provided an “automated image evaluation system” that accurately predicted the malignancy of lung nodules without any human intervention.

The researchers compared the AI’s evaluations with those of six board-certified U.S. radiologists who had up to 20 years of clinical experience.

When prior LDCT scans were not available, the AI “model outperformed all six radiologists with absolute reductions of 11% in false positives and 5% in false negatives,” report Tse and colleagues. When previous imaging was available, the AI performed just as well as the radiologists.

Study co-author Dr. Mozziyar Etemadi, a research assistant professor of anesthesiology at Northwestern University Feinberg School of Medicine in Chicago, explains why AI can outperform human evaluation.

“Radiologists generally examine hundreds of 2D images or ‘slices’ in a single CT scan, but this new machine learning system views the lungs in a huge, single 3D image,” Dr. Etemadi says.

AI in 3D can be much more sensitive in its ability to detect early lung cancer than the human eye looking at 2D images. This is technically ‘4D’ because it is not only looking at one CT scan but two (the current and prior scan) over time.”

Dr. Mozziyar Etemadi

“In order to build the AI to view the CTs in this way, you require an enormous computer system of Google-scale,” he continues. “The concept is novel, but the actual engineering of it is also novel because of the scale.”

Dr. Etemadi goes on to extol the benefits of using deep-learning technology, emphasizing its precision. “The system can categorize a lesion with more specificity,” the researcher says.

“Not only can we better diagnose someone with cancer, we can also say if someone doesn’t have cancer, potentially saving them from an invasive, costly, and risky lung biopsy,” concludes Dr. Etemadi.

The researchers caution, however, that it is first necessary to validate these results in larger cohorts.