KNOCKOUT AUTO ENCODER FOR DETECTING ANOMALIES IN BIOMEDICAL IMAGES

摘要

A mechanism is provided in a data processing system comprising a processor and a working memory, the working memory having instructions executed by the processor, in particular to configure the processor so that a knockout autoencoder control component for detecting anomalies in biomedical images is realized. The mechanism trains a neural network that is to be used as a knockout auto-encoder that predicts an original image on the basis of an input image. The knockout autoencoder control component provides a biomedical image as an input image for the neural network. The neural network outputs a probability distribution for each pixel in the biomedical image. Each probability distribution represents a predicted probability distribution of expected pixel values for a particular pixel in the biomedical image. An anomaly detection component executed within the knockout autoencoder control component determines a probability that each pixel has an expected value based on the probability distributions in order to form a plurality of expected pixel probabilities. The abnormality detection component detects an abnormality in the biomedical image based on the plurality of expected pixel probabilities. An anomaly marking component executed within the knockout autoencoder control component marks the detected anomaly in the biomedical image to form a marked biomedical image.