Researchers from NYU School of Medicine’s Technology have announced the first operational AI-enabled radiology assistant. The assistant can examine the report of the patients and predict the chances of occurrence of breast cancer. The development of the new platform is the result of the growth in the number of breast cancer cases. However, the major challenge lies, in the number of patients that have pseudo symptoms of malignancy. These patients have to undergo a series of tests to determine the state of their possible tumor post mammography. Nonetheless, these tests are extremely costly and require too much time for precise results.
What Made Introduction of AI so Necessary?
According to Dr. Krzysztof J. Geras, there was a high demand for a solution that can reduce the number of additional imaging in radiology. Dr. Geras also mentions that are chances when a radiologist might skip some cancer. Generally, they are too small to be detected. However, may cause serious problems in the future. With the new tool, the radiologist can feed the report to AI and it can analyze the patterns and determine wheater cancer will develop or not.
How Scientists Trained the Model?
In order to train the model to deliver precise and optimal results, the scientists fed the model with more than 800,000 data samples. These samples provide necessary guidance to the AI. However, the model is not yet ready to get operational commercially. This is because the state of cancer differs from patient to patient. Moreover, to use the model effectively, more data is constantly being fed. Hence, there are no plans to deploy the model any time soon.
Nevertheless, a pilot project, for proof of concept is under progress in NYU School of Medicine’s Technology. Here, the tests and images are looked into by radiologists as normally they do. If they are not sure about the probability of the occurrence of cancer, they might ask the AI model.