Ph.D. student Haipeng Li and his professor Ramakrishnan Mukundan have developed new computerised algorithms that help radiologists read mammograms.
The algorithms have been shown to detect two underlying factors linked to the increased risks of breast cancer.
Breast cancer is one of the most common cancers in New Zealand with one in nine Kiwis being diagnosed every year.
Recent Ministry of Health reports show nearly 600 women die every year from the disease and more than 3000 women will be diagnosed.
Maori and Pasifika women are 21 percent more likely to be diagnosed, 30 percent less likely to be diagnosed early, and 72 percent more likely to die.
The survival rate for women with the disease is 88 percent if treated early, so Li’s algorithms would help increase this number due to early detection.
Li said early detection through mammograms was an important part of preventing women’s deaths due to the disease.
“The algorithms I have designed are to make it easier for radiologists to pick up two markers, which are breast density and tiny calcium deposits."
Li’s new algorithms using artificial intelligence have shown a higher accuracy rate in detecting microcalcifications and denser breast tissue and provide a more specific location of the deposits.
He tested more than 700 images for density and managed to get an accuracy rating of 92 and 87 percent from two public datasets.
Li and his professor are aiming higher to get an overall accuracy of 98 percent which is why more tests are to be conducted.
His main goal is to provide a second opinion for radiologists as it would make their jobs easier and save a lot of women’s lives.
“Knowing that I am doing work that will contribute to people’s lives makes this research more meaningful to me."
BreastScreen Aotearoa aims to provide mammograms to 70 percent of New Zealand women aged between 45 and 69.
UC Computer Science and Software Engineering Professor Mukundan said he was thrilled with the work both Li and himself have produced over the last few months.
“Li’s research makes significant contributions to this field by performing texture analysis of mammographic features for microcalcification detection and breast density estimation."
Mukundan and Li hope this research and their algorithms will eventually make their way into radiologist's systems so they can identify cancers at an early stage and reduce the death rate.