MIT Computer Science and Artificial Intelligence Lab: "AI can predict breast cancer five years in advance, as well as white and black patients."
The Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory has developed a new deep learning-based artificial intelligence prediction model that predicts breast cancer development five years ahead of schedule. In the course of development, the researchers realized that other similar artificial intelligence prediction models usually carry human-specific biases, because those projects are mostly designed based on white patient populations. For this reason, the MIT research team designed a new artificial intelligence prediction model to use "fairer" data to ensure that white and black women's predictions are equally accurate.
The research team pointed out in a blog post that black women are more than 42% more likely to die of breast cancer than white women. One factor may be that current early detection techniques do not provide adequate services for black women. The team said that the purpose of the development of the testing technology was to target ethnic minorities so that they could know whether they would "bring health risks and risks due to breast cancer." From the overall development of artificial intelligence in the past, due to algorithmic bias, these ethnic minorities are not well represented in developing deep learning models. Algorithmic bias is also the focus of many industries in the current research of artificial intelligence, and many AI companies are focusing on improving this when they launch new products in different fields.
The predictive model completed deep learning of more than 60,000 patients (over 90,000 mammograms in total) in Massachusetts General Hospital to receive mammograms and patient prognosis (final cancer development is critical). Begin with data screening analysis and use deep learning recognition patterns to identify lesions that clinicians cannot directly observe. From the above experiments, the prediction model is based on actual x-rays rather than existing assumptions or known risk factors (these are at best a suggestive framework). So far, the experimental results of this predictive model have shown its accuracy, especially in terms of prediction and pre-diagnosis.
Overall, the program is designed to help medical professionals develop a correct screening plan for patients through computerized methods so that early results can be treated in time to avoid late diagnosis. The MIT research team hopes that the technology can also be used to improve other disease detections for similar problems with existing risk models (large gaps and low accuracy).
Source: WuXi PharmaTech AI
Smart Factory
Smart Factory
Automated Precision Inc. , https://www.apiasean.com