Ole Johan Skrede,Tarjei Sveinsgjerd Hveem,John Robert Maddison,Havard Emil Greger Danielsen,Knut Liestøl
申请号：
GB201718970
公开号：
GB2569103A
申请日:
2017.11.16
申请国别(地区):
GB
年份:
2019
代理人:
摘要：
A machine learning algorithm, optionally a neural network, or convolutional neural network, is trained on a plurality of histological microscopic images and a measure of outcome of each image – step 210. Each image is divided into tiles step 220. The measure of outcome is assigned to each tile of the image. The tiles are then used to train the machine learning algorithm. The trained algorithm may then be used to evaluate images stained with a marker, obtained using at least two different pieces of equipment. The microscopic image may be a colour image consisting of one, two or three colour channels. The training may further comprise determining a first threshold value and a second threshold value, counting the number of scores above or below the first threshold value, and comparing the counted number of scores with the second threshold value to obtain a binary single predicted outcome value depending on whether the counted number of scores is above or below the second threshold； and optimizing the first and second threshold values to minimize the variance between the binary single predicted outcome value and the respective measure of outcome for the respective measure of outcome for the plurality of images. A representation of the scores may be output as an image of tiles, each tile represented by a grey colour corresponding to its score.
