New Delhi: Researchers have developed a new deep learning model that can estimate breast density, which could be useful for the prediction of cancer risk. Researchers at the University of Manchester, UK, said the automatic feature extraction from training data enabled by the deep learning-based approach makes it attractive to breast density estimates
Breast density is defined as the proportion of fibroglandular tissue inside the breast and is often used to assess the risk of developing breast cancer.
They have published their findings in the Journal of Medical Imaging.
The researchers used two independent deep learning models, initially trained on ImageNet, a non-medical imaging dataset with over a million images, and trained them on their medical imaging data through an approach known as “device learning.” transfer”.
Training and building deep learning models from scratch is challenging due to limited data sets, they said.
Experts including radiologists, advanced professional radiographers, and breast physicians assigned density values on a visual analog scale in 1,60,000 complete fields. digital mammography images they form 39,357 women.
Using this data, the researchers developed a procedure that would estimate a density score by entering a mammography image as input.
The procedure involved pre-processing the images to make the training process less computationally intensive, extracting features from the images processed with the deep learning models, assigning the features to a set of density scores, and then combining the scores using a joint approach to produce a final result. density estimation.
Thus, the team developed highly accurate models to estimate breast density and its correlation with cancer risk, conserving calculation time and memory.
“The performance of the model is comparable to that of human experts within the limits of uncertainty,” said principal investigator Susan M. Astley. “In addition, you can train much faster and on small data sets or subsets of the large data set.”
The researchers noted that the framework is not only limited to estimating breast cancer risk, but also training other medical imaging models based on their estimates of breast tissue density.
Breast cancer is the most common cancer affecting women worldwide.
Although there are several methods available to estimate this measure, studies have shown that subjective assessments by radiologists based on visual analogue scales are more accurate than any other method.
