Picture of Dr. El-Shenawee and Dr. Nelson

Engineering Professors to Develop Technology Aimed to Fight Breast Cancer


Two University of Arkansas engineering professors received a $19,145 grant from the University of Arkansas Women’s Giving Circle to develop technology that could help fight breast cancer.

The award was given to Magda O. El-Shenawee, professor of electrical engineering, and Alexander Nelson, assistant professor of computer science and computer engineering.

The goal of their research is to develop technology that will help doctors identify the type of cancer patients may face – whether life threating or not. Identifying the disease from the first stage, known as ductal carcinoma in situ, or DCIS, could prevent patients from enduring harsh treatments like chemotherapy.

Breast cancer became an interest for El-Shenawee two years ago when she and her former student Tyler Bowman met with a breast surgeon presenting terahertz imaging for margin assessment. During their visit, the surgeon expressed that it would be helpful to him and his patients if they could help identify the stage of DCIS cancer.

“He said this would be more significant than the margins,” El-Shenawee said.

In breast-conserving surgery – an alternative to a mastectomy – margins are the rim of normal tissue surrounding a tumor that is removed during surgery. If a margin is too large, it removes healthy tissue, or if it’s too small, it may indicate a chance that some cancer was left in the breast. Analyzing margins through pathology takes weeks while the patient waits for results. El-Shenawee’s terahertz technology can potentially reduce this time, aiming at real-time.

El-Shenawee and Nelson want to explore terahertz-imaging technology to improve patients’ quality of life. They want to develop a reliable biomarker of DCIS using terahertz imaging technology and a deep machine-learning methodology based on convolutional neural networks to classify DCIS lesions along non-invasive and invasive grades.

DCIS represents one of the most controversial diagnoses, according to the U of A researchers.

“Some breast physicians would consider DCIS non-invasive and needs no further treatment, while others consider it invasive and need aggressive treatment,” El-Shenawee said.

DCIS is non-invasive breast cancer because it does not spread outside the “milk duct,” however; this disease could cause future problems, according to BreastCancer.Org.

“When you have had DCIS, you are at higher risk for cancer coming back or for developing new breast cancer than a person who has never had breast cancer before,” BreastCancer.Org website states.

Nelson said machine learning is the process of algorithms “learning” answers to problems. This field of study has exploded due to increasing international focus on data. Deep learning, a subset of machine learning, leverages multiple hierarchical layers to extract higher orders of information. A 2017 article in Nature by Andre Esteva described a deep-learning system that could obtain a dermatologist-level classification of skin cancer from images. Nelson believes this technology could provide a tool to better understand pre-malignant cancers like DCIS.

“Deep learning in medicine isn’t just about developing algorithms and networks to classify. It is important to understand why those networks make those determinations. Terahertz spectroscopy allows for a second view that can be combined with traditional pathology to perform these analyses,” Nelson said. 

Researchers hope that their preliminary analysis will lead to new findings before the end of the year. 

“We won’t be able to completely answer the underlying mechanisms with our current amount of data. But we hope to have a better understanding of DCIS with our methodology, which could lead to higher-level questions and ultimately to improved quality of life for millions of women,” Nelson said.