AI assists in lung cancer diagnosis

Original article was published on Artificial Intelligence on Medium


AI assists in lung cancer diagnosis

In recent years, AI has been gradually applied in the medical field, especially in the diagnosis of diseases such as lung cancer.

AI helps to improve the early diagnosis accuracy rate of lung cancer.

Lung cancer is the most common malignant tumor in the world, with morbidity and mortality ranking first in malignant tumors, and has become a recognized killer of human health. The prognosis of lung cancer is closely related to the clinical stage. Due to the late appearance of symptoms and signs, most patients have metastasized at the first visit to the hospital, and the 5-year survival rate is only 16% due to the missed optimal surgery time. Therefore, clinical staff are constantly looking for newer and more sensitive imaging techniques suitable for lung cancer screening.

In August 2002, the National Lung Screening Trial (NLST) led the launch of a randomized controlled clinical trial comparing lung cancer screening with low-dose spiral CT (LDCT) and ordinary chest X-rays, which is by far the most authoritative in the world. However, the NLST study also found that only 0.6–2.7% of patients with lung nodules found through the clinical screening practice of low-dose spiral CT were eventually diagnosed with lung cancer.

In the traditional method of early diagnosis of pulmonary nodules, simple imaging data requires long-term radiological follow-up of the patient to observe the imaging morphological changes, which may bring about potential radiation damage, while invasive diagnostic operations or direct surgical treatment may cause physical and psychological damage to patients. Thanks to the rapid development of novel liquid biopsy and AI diagnosis, researchers have seen a revolutionary change in the early diagnosis of pulmonary nodules.

The new diagnosis model “Biomarker discovery + AI”

The use of AI medical image analysis to assist doctors in screening for esophageal cancer, lung nodules, diabetic retinopathy, colorectal tumors, breast cancer and other diseases has been realized. In total, AI assisted diagnosis can help assist doctors in identifying and predicting the risk of more than 700 diseases. In the identification of lung nodules, researchers can use computer vision and deep learning technology to assist doctors in reading through medical image, accurately locating tiny lung nodules more than 3mm, and determining whether the nodule is benign or malignant.

How to further assist in improving the diagnostic efficiency of AI?

The liquid biopsy technology today can detect and trace biological markers released into the blood by early tumors, such as microRNA, circulating tumor DNA, circulating tumor cells, etc. It is believed that the combination of liquid biomarker biopsy and AI technology will inevitably improve the accuracy of early lung nodule diagnosis. As expected, in the ABC model of compound clinical features (Clinic), biomarkers (Biomarkers) and artificial intelligence results (AI), the effectiveness of diagnostic tools is as high as 0.955 (1 indicates that the highest diagnosis efficiency). In the subsequent verification group, the ABC model also showed higher area value and sensitivity than other models, which means that the diagnostic model of “biomarker + AI” can be more accurate.

Since the biomedical imaging technology is not mature enough yet, the construction of a multi-modal “biomarker + AI” is the ideal mode for clinical pulmonary nodule diagnosis at this stage.

AI-assisted diagnostic technology must have a leap forward in the future.

AI has brought a huge change to the traditional way of classifying and management of imaging data. It can process tens of thousands of image information quickly and simultaneously, which will greatly save the physical and mental effort of professional imaging doctors. With the development of biological imaging technology, AI-assisted diagnostic technology will certainly achieve fast progress.

AI medicine is a brand-new field that integrates medicine and industry. Driven by big data, artificial intelligence, cloud computing and other technologies, AI’s role as a medical assistant has made the medical community full of expectations. The cross-border integration of medical and technology is continuously promoting the application of AI medical imaging, which effectively connects AI and application scenarios.