AWARrD Project 1


aim


Artificial Intelligence for the Myometrium

Uterine Leiomyomas or Fibroids - Common, but how do we know if they are cancers?

By the age of 50 at least 70-80% of African women will develop tumors in their uterus that are called leiomyomas, or, more commonly, fibroids. In the vast majority of instances, fibroids never cause any symptoms at all, and frequently, when a woman experiences symptoms of abnormal vaginal bleeding, pressure or even infertility, fibroids that she might have are not associated with the problem facing her. However, in some instances, fibroids can cause heavy menstrual bleeding (HMB) - sometimes without even enlarging the uterus, or they may contribute to infertility or recurrent pregnancy loss, and, if they become very large, they can manifest with symptoms related to pressure on surrounding organs and structures. While this is an issue in the foreground, there is another lurking issue and that is the risk of the fibroid actually being a cancer - a leiomyosarcoma. Compared to benign or non-malignant fibroids, leiomyosarcomas are quite rare, but when you consider the frequency of leiomyomas, it is a consideration. There are at least three unfortunate issues. One is that we don't have a good idea about the frequency of malignancy - estimates range from about 1/250 to 1/7,000 generally being higher in women in their fifth and sixth decades. The second problem is that leiomyosarcomas are associated with a high death rate - it is difficult to successfully treat these tumors before they have spread to other organs. The last issues is that we don't, as of yet, have a way to diagnose leiomyomas that are suspected to be malignant. So that is where the AWARrD team and data science comes in.

Artificial Intelligence for the Myometrium - Teaching machines to detect leiomyosarcoma

The aim project is designed to test the hypothesis that by exposing magnetic resonance images (MRIs) of knowingly benign and malignant leiomyomas - leiomyosarcomas - to a "deep learning" computerized system, the "machine" may "learn" to diagnose the leiomyosarcomas by detecting subtle differences that may escape the human eye. For the aim project, a database of  hundreds of MRIs of benign fibroids will be used to train the machine on recognizing benign tumors, and a set of MRIs of leiomyosarcoma will be found to show the system the malignant version. Computer scientists with expertise in "deep learning" at the University of Assiut and University of Chicago will develop this project and hopefully will help us find a way to diagnose these rare but dangerous tumors.

Share by: