Machine Learning and Lung Cancer Detection

The Institute of Medicine at the National Academies of Science, Engineering and Medicine reports that diagnostic errors contribute to approximately 10 percent of patient deaths and up to 17 percent of hospital complications.

According to Xin Zhong, co-founder and CEO of Sigma Technologies, the traditional diagnostic approach often detects the cancer too late, leading to a 17 percent survival rate for patients. An AI algorithm system is working towards reducing diagnostic errors associated with lung cancer in its early stages.

Traditionally, doctors diagnose lung cancer by examining CT scan images and inspecting small nodules to identify whether the nodules are benign or malignant. Small nodules often remain unnoticed because of their size and the expertise required to find, recognize and properly classify the nodules.

It takes 10 minutes for a doctor to visually inspect the patient’s CT images for nodules, with additional time needed to classify the nodules as benign or malignant. There is a high risk of human error. 12 Sigma’s AI algorithm inspects the CT images and classifies any nodules within a two minutes. Zhong estimates that the system saves doctors four hours per day, allowing more time to focus on patients, further expand their skill set, or other duties.

The AI system works alongside the doctors in 35 hospitals in China. As doctors help fine-tune the system by removing inaccurate data, the AI algorithm system becomes more accurate and helps decrease the time required for a diagnosis.

This is just one example of machine learning in the medical field. Researchers at Stanford have trained an algorithm to diagnose skin cancer, while Harvard is using deep learning to help interpret pathology images.