Sunday , June 26 2022

Google researchers have created an algorithm to detect lung cancer with greater certainty than a radiologist


Artificial intelligence researchers who work in the Google giant have developed a high-performance diagnostic algorithm, saving lives literally. After its publication on Monday, the study revealed a deep learning algorithm that can detect, through analysis of MRI scans, lung cancer with a success rate of 94.4%!

According to the study's results, as well as its high success rate, the algorithm performs better than radiologists under certain circumstances. The system has achieved this success rate on 6716 cases during the National Lung Cancer Screening Trial, with similar detail on 1139 independent clinical cases.

The researchers carried out two separate studies, one where screening was available for analysis and one where screening was not available. In the first scenario, the deep learning algorithm showed a higher identification rate than six radiologists, reported the New York Times.

To achieve such performance, the algorithm was formed by analyzing CT scans data of people with lung cancer or cancerous nodules, comparing them with some healthy subjects.

" The whole experiment process looks like a student in school Daniel Tse, Google Project Manager, told New York Times. " We use a large set of data for learning, giving you lessons; and quizzes so that they can begin to learn for themselves and understand what cancer is, or what will or will not become cancer in the future. We then gave him a final examination based on data that he had never seen from the front, having spent a lot of time learning. The result obtained in the final examination: a! ".

You may also like: Babylon: artificial intelligence that diagnoses patients as well as experienced doctors t

But in fact, although promising, the results of this study only show the first stages of algorithmic detection of diseases. The system is still untested and is accurate enough to be increased in all healthcare settings offering cancer screenings. But this version is already promising to automate the fight against false positives and negative negatives.

" The lung scan for smokers is so bad that it is difficult to make things worse t Eric Topol, director of Scripps Research Translational Institute in California said.

Many technology companies, including Google, already rely on algorithms such as detection tools on their platforms, and this is already large-scale (for example for standardizing online communities). But these automated systems are still precise, leading to false censorship or inability to identify violent and hate content, which sometimes spread very quickly.

Researchers working on lung cancer detection technology recognize the dangers and risks of setting up such a system. They are aware of the need to establish a management system and above all, as well as validating its effectiveness.

" We are working with organizations around the world to understand how this technology can be practically implemented in clinical practice Dr. Tse told the New York Times. " We do not want to get on with ourselves ".

Source link