Apr 02, 2020 12:06:00

Research results that can easily screen more than 50 types of cancer with new blood test by AI

AI (artificial intelligence) using machine learning algorithms is already expected to be applied to medical settings, and it is possible to detect lung cancer from CT scan data with higher accuracy than human doctors and to be effective with cancer immunotherapy. There have been reports of such things as determining who does not work . Research teams in the United Kingdom and the United States have announced that a new blood test using AI can easily screen more than 50 types of cancer.

Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA-Annals of Oncology
https://www.annalsofoncology.org/article/S0923-7534(20)36058-0/fulltext

New blood test can detect wide range of cancers, now available to at risk individuals in clinical study at Dana-Farber-Dana-Farber Cancer Institute | Boston, MA
https://www.dana-farber.org/newsroom/news-releases/2020/new-blood-test-can-detect-wide-range-of-cancers--now-available-to-at-risk-individuals -in-clinical-study-at-dana-farber /

New AI-Powered Blood Test Promises Easy Screening For Over 50 Types of Cancer
https://www.sciencealert.com/ai-powered-blood-test-promises-quick-and-easy-screening-for-over-50-types-of-cancer

New blood test can detect 50 types of cancer | Science | The Guardian
https://www.theguardian.com/science/2020/mar/31/new-blood-test-can-detect-50-types-of-cancer

Most conventional blood tests for cancer diagnosis detect the DNA of cancer cells that flow into the blood when immunity destroys the cancer cells, based on

DNA sequencing that determines the base sequence of the DNA. did. However, the research team has focused on the DNA methylation pattern of cancer cells instead of the DNA base sequence and developed a new blood test approach. Analyze DNA methylation patterns in blood when cancer cells are destroyed by immunity and flow into the blood because the methylation pattern occurring in the DNA of cancer cells is significantly different from normal cells. Can be used to determine the presence or absence of cancer.

Dr. Jeffrey Oxnard, a member of the Dana-Farber Cancer Institute , Harvard Medical School's primary health care provider, said, `` Pregnant women float in the blood to check for abnormalities in the fetus. Examining DNA: We know that there are approaches based on DNA in the blood, but the question is how to fine-tune and perfect the technology for finding cancer from DNA in the blood Is it? ' So the team developed a new blood test approach using machine learning algorithms.

First, the team took blood samples from more than 2,800 cancer patients and trained the algorithm on the DNA methylation patterns in the samples. Next, he said that he trained the algorithm using DNA methylation patterns contained in 3052 blood samples, including 1531 patients diagnosed with cancer and 1521 non-cancer patients. The research team told the algorithm not only about the presence or absence of cancer, but also about 'where the patient in this sample has cancer'.

The team then tested the trained system with blood samples from 1264 people, with cancer about half of the total. The sample included more than 50 types of cancer, including breast, esophageal, stomach, ovarian, lung, multiple myeloma, and pancreatic cancer.

As a result of the test, the new system seems to correctly detect cancer in 44% of blood samples of people diagnosed as having cancer, especially saying, The research team says the percentage was very low, 0.7%. In addition, the accuracy of cancer detection seems to have improved as the cancer progressed, with 18% of all stage I cancers being detected and 93% of stage IV cancers being detected. Was.

In some cancers, the detection accuracy was particularly high. For pancreatic cancer, 63% was detected in stage I and 100% in stage IV. In addition, the system could not only detect the presence of cancer, but also identify the type of cancer. In 96% of cases where the system determined that there was cancer, the system guessed where the cancer was, and this guess was accurate with 93% accuracy.

The research team of Dr. Oxnard and others has developed a blood test that can detect more than 20 types of cancer in 2019, but it seems that the accuracy is further improved by a system using a machine learning algorithm.

A blood test that can identify more than 20 types of cancer will be developed-gigazine

According to Dr. Oxnard, the system is already undergoing clinical trials. The results showed that cancer detection by AI-based blood tests is promising.However, due to the problem of low early cancer detection rates, `` This blood test is routinely used. Before doing so, you will need to look at the results of clinical trials to better understand the performance of the test, 'said Oxnard.