Study: AI cancer screener is 98% accurate
A new artificial intelligence (AI) model is able to detect 13 different types of cancer with 98.2% accuracy, but only using DNA data from tissue samples, new research shows. The model, called EMethylNET, was developed by researchers at the University of Cambridge in the UK and has the potential to accelerate early detection, diagnosis and treatment of cancer.
The findings, published last week in the journal Biology Methods and Protocols, focus on DNA methylation, a chemical process that occurs when cells begin to grow, including cancer cells. Researchers trained AI models to identify early-stage cancer structures and pathways.
The research report states:
"Cancer is a group of more than two hundred different diseases and remains the leading cause of death in humans. Cancer is typically detected in its terminal stages, and metastatic cancers account for 90% of cancer-related deaths."
For this reason, the accuracy of early cancer screening is very important and will have a significant impact on the survival rate and treatment of cancer.
Image source: Medical Device Network
Cambridge trains AI model to detect 13 types of cancer early
How an AI cancer screener works
How do researchers train AI? First, they trained EMethylNET on data from the Cancer DNA Genome Atlas on more than 6,000 tissue samples from 13 cancer types, including breast, lung and colorectal cancer. They then tested it on more than 900 samples from an independent data set.
The most important finding was that it was more than 98% accurate in classifying 13 cancer types and non-cancer samples. The study also highlights that AI performs well even when examining diverse data from different countries. The researchers were also able to identify 3,388 methylation sites associated with cancer-related genes and pathways.
According to the study, the model combines two AI methods: using XGBoost to select relevant features and using deep neural networks for classification. This allows AI to not only accurately detect cancer, but also provide insights into the body's regulation of non-genetic factors that mutate normal cells into cancer cells.
"These epigenetic modifications are some of the earliest tumor events associated with cancer development," the study noted, highlighting the potential of this AI in early cancer screening.
Image source: academic.oup
Cambridge AI model cancer screening research paper
Don't be too optimistic yet? Researcher: More testing is needed
While the preliminary findings look very promising, the study authors caution that further research and testing is needed before it can be used clinically. The research team said they are currently working to apply the model to liquid tissue samples, which could be used to provide non-invasive early cancer screening.
Even though more testing is needed, the researchers' ultimate goal is to have EMethylNET screen for hundreds of cancer types. As AI continues to make its way deeper into the medical field, using AI models for earlier and more accurate cancer diagnosis could have a profound impact on public health.
Cancer remains the leading cause of death in Taiwan
Cancer is the number one enemy of human health. According to statistics from Taiwan's Ministry of Health and Welfare, among the top ten causes of death in 2023, malignant tumors (cancer) still rank first, followed by heart disease, pneumonia, and cerebrovascular diseases.
According to statistics, 53,126 Taiwanese will die from cancer in 2023, accounting for 25.8% of the total deaths. The mortality rate will increase by 2.2% compared with 2022. The top ten most lethal cancer types are: lung cancer, liver cancer, rectal cancer, breast cancer, etc.
However, as technology continues to advance and more data is accumulated, AI models may become an important tool in cancer screening and change the way humans fight cancer.