Aurora testing the ability of its Deep Learning Engine
Aurora recently had their research engineers apply its Deep Learning platform — an artificial intelligence (AI) system that trains a computer’s memory to recognize unique patterns of characteristics by processing large amounts of data — to recognize the difference between cancerous and non-cancerous tissue.
Using a dataset of histopathology images that were pre-labelled as either containing cancer or not containing cancer, the DL platform was able to recognize the difference between the cancerous and non-cancerous tissue.
The histopathology images were placed under a microscope and stained to enhance cellular features in the image.
The detection of mitosis (cells that are in the process of division) is one of the most popular methods for diagnosing breast cancer on histology images.
Physicians will typically analyze each individual frame and mark the presence of mitosis, which can be a cumbersome and time consuming process.
As such, there has been a growing interest in developing an accurate and automatic solution to this issue. The International Conference of Pattern Recognition launched a medical imaging
challenge in 2014 that focused on finding a proper solution to address this issue.
Using the dataset provided for the challenge, Aurora developed an AI to detect the occurrence of mitosis.
In a blind test set, the AI demonstrated outstanding results and even suggested a higher level of accuracy than the 2014 competition winner.
Based on these results, Aurora contends that its Deep Learning platform has the power and capability to be applied to completely unrelated fields, in addition to its main focus in face recognition.
Previously reported, Aurora announced that the University of Nottingham Computer Vision Lab has independently verified Aurora’s facial recognition technology as one of leading solutions in the field.