COLLEGE PARK, Md.?In movies like The Terminator and The Matrix, evil forms of artificial intelligence attempted to enslave mankind. Researchers at the University of Maryland’s Greenebaum Cancer Center are proving artificial intelligence can do good things, too.
Using Artificial Neural Networks, researchers developed and tested a method of identifying and diagnosing different types of colon cancer?the third most common cancer. One of the goals of the study is to use ANNs for early cancer diagnosis, sparing patients from unnecessary surgeries and in many cases saving their lives, said Florin Selaru, the study’s author and a research associate at the UM School of Medicine. Selaru is also the director of bioinformatics data analysis at the Greenebaum Cancer Center.
The study used 39 tissue samples from colon cancer patients at the Baltimore VA Medical Center and Mount Sinai Hospital in New York.
Researchers extracted genetic material from each of the samples and used microarray equipment to analyze the 8,064 genes in each sample. The gene expression, or the level at which each gene is represented, was measured for each colon sample and placed into spreadsheets.
The microarrays make note of abnormal expression levels, which are often related to cancer. These genes could cause cancer or could be symbolic of cancer.
The amount of gene expression data in the spreadsheets is huge, Selaru said.
“We don’t have the ability as humans to look at the data,” he said.
Enter ANNs: ANNs need large amounts of data to analyze, Selaru said. Working like the human brain, ANNs can examine and learn from data.
The genes for 27 of the samples were encoded to represent cancer. The information was fed into the ANNs, which examined the information and produced an analysis.