99.26 % accurate: Researchers reveal new AI for cancer detection

The gold rush for the most accurate cancer detection AI appears to have a new leader for now.

Researchers have designed a new specialized artificial intelligence to detect cancer in microscope images that achieves a success rate of 99.26 % for diagnosing Endometrial cancer, which is a new milestone.

They are now looking for other cancers to discover.

The participants in the study were Daffodil International University in Bangladesh, Charles Darwin University, the University of Calgary and Australian Catholic University.

Far better than the human eye
The average human-led detection rate is 78.91 % to 80.93 %, making this new technique significantly better than than any given doctors eyeballs.

The AI can apply magnification on problem areas and run the images through a series of filters for easier detection, with a precision almost undetectable to the human eye.

A very common cancer
Endometrial cancer is the most common cervical cancer, and affects 320 000 women globally each year, according to Wikipedia.

It affects about 1.6 % of woman worldwide in their lifespans, and early detection, like with this tool, is crucial to successful treatment.

The researchers are not intending to replace humans in diagnosis with this tool, but foresee a world where AI assistants work with doctors to make detection more accurate, so treatment can start sooner.

Also effective on other cancers
They are also upbeat that their tool, called ECgMLP, can help diagnose other cancers:

— The same methodology can be applied for fast and accurate early detection and diagnosis of other diseases which ultimately leads to better patient outcomes, Associate Professor Shafiabady at Australian Catholic University said in a press release.

They are not just thinking about detecting other cancers with high accuracy, but are actively pursuing this avenue with similar accuracies:

— We evaluated the model on several histopathology image datasets. It diagnosed colorectoral cancer with 98.57 per cent accuracy, breast cancer with 98.20 per cent accuracy, and oral cancer with 97.34 per cent accuracy.

Those numbers, too, would be better than the human eye of a well trained specialist.

These researchers are hadly alone in this field, however. There are efforts by Google that also claim 99 % efficiency in breast cancer, and there’s Harvard claiming 94 % in many cancers, just to mention a few.

Read more: Charles Darwin University announcement, A report on New Atlas and The actual scientific paper.