Scientists at Google and Northwestern Medicine recently determined the precision of new deep learning systems for predicting lung cancer. The deep learning tools were able to detect malignant lung nodules on low-dose chest computed tomography (LDCT) scans with a performance that met or exceeded what a team of expert radiologists found.
These deep learning systems provide an automated image evaluation system for enhancing the accuracy of early lung cancer diagnosis that could potentially lead to earlier treatment. And the deep learning system was compared against radiologists on LDCTs for patients — some of whom had biopsy confirm cancer within a year.
In most comparisons, the deep learning model performed at or better than radiologists. And the deep learning system produced fewer false positives and fewer false negatives — which could lead to fewer unnecessary follow-up procedures and fewer missed tumors in clinical settings. This research was published in Nature Medicine on May 20.
“Radiologists generally examine hundreds of two-dimensional images or ‘slices’ in a single CT scan but this new machine learning system views the lungs in a huge, single three-dimensional image,” said study co-author Dr. Mozziyar Etemadi — a research assistant professor of anesthesiology at Northwestern University Feinberg School of Medicine and of engineering at McCormick School of Engineering. “AI in 3D can be much more sensitive in its ability to detect early lung cancer than the human eye looking at 2-D images. This is technically ‘4D’ because it is not only looking at one CT scan, but two (the current and prior scan) over time.
Etemadi leads a research team while also working in anesthesiology residency training at Northwestern as part of a unique residency research track. And Etemadi’s dual roles enables research in his lab to combine healthcare and advanced engineering. Etemadi’s lab is based in one of the intensive care units at Northwestern Memorial Hospital — which allows seamless communication among engineers and nurses, physicians, and other care providers.
“This area of research is incredibly important, as lung cancer has the highest rate of mortality among all cancers, and there are many challenges in the way of broad adoption of lung cancer screening,” added Google technical lead Shravya Shetty. “Our work examines ways AI can be used to improve the accuracy and optimize the screening process, in ways that could help with the implementation of screening programs. The results are promising, and we look forward to continuing our work with partners and peers.”
Currently, lung cancer is the most common cause of cancer-related death in the U.S. as it resulted in 160,000 deaths last year. And large clinical trials across the US and Europe have determined that chest screening can identify cancer and reduce death rates. But high error rates and the limited access to screenings mean that many lung cancers are usually detected at advanced stages when it is hard to treat.
The deep learning system utilizes both the primary CT scan and prior CT scan whenever available from the patient as input. And the prior CT scans are useful in predicting lung cancer malignancy risk since the growth rate of suspicious lung nodules can be indicative of malignancy.
The computer was trained using fully de-identified and biopsy-confirmed low-dose chest CT scans. And this novel system identifies both a region of interest and whether the region has a high likelihood of lung cancer. This model outperformed six radiologists when previous CT imaging unavailable and performed as well as the radiologists when there was prior imaging.
Etemadi pointed out that the system can categorize a lesion with more specificity. “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,” explained Etemadi.
Google’s scientists developed the deep learning model and applied it to 2,763 de-identified CT scan sets provided by Northwestern Medicine in order to validate the accuracy of its new system. Google product manager Dr. Daniel Tse was the corresponding author on the paper.
And the scientists found the AI-powered system was able to spot sometimes-minuscule malignant lung nodules with a model AUC of 0.94 test cases. These cases were pulled from the Northwestern Electronic Data Warehouse along with other Northwestern Medicine data sources.
