A new paper published on Arxiv.org by researchers from the University of Central Florida Centre for Research in Computer Vision has detailed a new deep learning method of detecting lung cancer in patients. So far, the researchers have been able to detect lung cancer nodules with up 97% accuracy.
Lung cancer is one of the most aggressive forms of cancer. The American Cancer Society estimates more people die every year from lung cancer in the US (154,000 people) than die from breast, colon and prostate cancer combined. Lung cancer is also extremely time sensitive with only 4% of newly diagnosed patients living for five years if the cancer has been identified after it has already spread.
Hence, the ability to improve detection rates through AI is a welcomed advancement. The researchers used convolutional neural networks capable of mimicking the way the human brain's neurons react with one another, to target the small nodules that comprise lung cancer tumors with a sensitivity of between 95-97%.
In the report, the researchers state they "used publicly available … [scans] and showed that the proposed method outperforms the current literature both in terms of efficiency and accuracy …To the best of our knowledge, this is the first study to perform lung nodule detection in one step."
"A promising future direction will be to combine diagnosis stage with the detection," the team later adds.