Enhanced accuracy and reliability of ER and PR immunohistochemistry scoring using ANN from digital microscope images

Abstract EN

ER (Estrogen receptor) and PR (Progesterone receptor) are breast oncogene receptors that are important for the growth of some organs.

Their statuses for breast cancer patients are vital to determine chemical therapy for the patients after surgical removal of cancer.

A computer program programmed by the author in Matlab language to study each type of receptors (10 slide images studied for each one).

The program used pixel color classification techniques by using Artificial Neural Network (ANN).

As a results: Some images resizing, re-dimensioning, and enhancing are done by photoshop program and that aimed to obtain better results of Proportion Observation percent (PO%).

The PO% estimated for ER and PR receptors by two methods; pathologist method (manual scoring), and by computer (computer scoring).

Results of the methods are close to each other for ER and PR receptors and statistical evaluation confirmed that the program can be used as an objective method to confirm the first method.

Computer program was better than the manual method for PO% computation because it is an objective method and shows results in very accurate manner.