- Google’s research team has come up with a deep learning model that can detect cancer
- It is applied to a microscope to detect this in real time
- The model was trained on images of human tissue and the testing results have been impressive, with the AUC as high as 0.98
Detecting cancer at an early stage has long been a focus area in healthcare. From IBM Watson to other major players, a lot of money has been spent trying to make headway in this field but with little success.
Now, Google has entered the crowded field and made a splash with their promising results. They have developed a deep learning model that has been incorporated in a microscope that doctors can use to detect cancer.
This is all very fascinating, but how does this AI actually work?
As with all deep learning studies, a deep neural network was trained to detect cancer cells by analysing images of human tissues. Then, a slide with human tissue is placed under the lens of the modified microscope. The image that you see in the microscope is sent to a computer and the deep learning model does to work detecting cancer in the tissue.
This is all done in real-time and is fast enough that when the slide is moved, the results are still populated in the computer.
The hardware setup is made up of a modified light microscope that enables real-time image analysis and presentation of the outcomes of the machine learning algorithms. The best part about this is that the microscope can be retrofitted in the existing setup in hospitals and clinics around the world, at a relatively low cost.
The initial results have shown a lot of promise. The researchers tested the model on two different cancer types – breast cancer and prostate cancer. A test on the former showed an AUC (area under the curve) as 0.98 and for the latter, the AUC was 0.96.
The study has also been jotted down in a research paper that you can access here and also view Google’s official blog post here. This is still pending review and the researchers have said they need to perform a more in-depth study to make the model far more accurate and to overcome the current limitations.
Check out Google’s video below depicting this AI:
Our take on this
Google has previously published results on how the team used a trained convolutional neural network to detect breast cancer. But this latest release, pending review, will be a boost in the arm for the healthcare community. It can potentially be used to detect other illnesses like tuberculosis and malaria.
These AR microscopes can be used beyond healthcare – in life sciences, manufacturing, and material research as well.
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
You can also read this article on our Mobile APP