Researchers from Stanford have developed an algorithm that can evaluate chest X-rays and analyse signs of diseases.
The research work is going at a good pace, and it has already tasted success. In only a month's time or more, these algorithms have produced better results than the radiologists at diagnosing pneumonia. Apart from pneumonia, it can also diagnose 14 other types of medical conditions.
According to an article at the official site of Standford University, CheXNet, a paper about the algorithms and its developments, was published this month.
The initial work started with the collection of over 112,000 frontal view chest X-ray images from the dataset released by National Institutes of Health Clinical centre. All these x-rays had around 14 different pathologies.
The research team, consisting of the Machine Learning Group, along with a few professors of radiology worked on ensuring that the algorithm's initial foundation was free from errors and gave consistent results.
The researchers acknowledged the challenges of interpreting X-ray images to diagnose severe pathologies like pneumonia. In many cases, radiologists get variations in the diagnoses. The research team planned on working on that gap and produce accurate diagnoses by building machine learning algorithms which could develop a database following the study of numerous diagnoses.
The paper submitted by the research team also includes the development of a computer-based tool which will highlight the areas that the algorithms identified as having higher chances of pneumonia. This device could help increase the efficiency of the radiologists as it will provide the hints of where to look first which will ultimately lead to quicker diagnoses for the ailing patients.