Researchers using artificial intelligence have developed a method to determine which types of skin cancers may be highly metastatic.
August 11, 2021- Researchers at UT Southwestern Medical Center have discovered a method that uses artificial intelligence to predict which types of skin cancers are highly metastatic. Cell Systems’ study looked at how AI-based tools could revolutionize the pathology of cancer and other diseases.
“We now have a general framework that enables us to take tissue samples and predict the mechanisms within cells that cause disease, mechanisms that are currently inaccessible in any other way,” said the study director and professor and chair of the department for Bioinformatics Lyda Hill at the UTSW Gaudenz Danuser, PhD, said in a press release.
AI technology has developed a lot in recent years. According to Danuser, using methods based on deep learning, AI technology can distinguish differences in images that are invisible to the human eye.
Researchers have recommended the use of artificial intelligence to look for differences in disease characteristics, to provide information about diagnoses or to guide treatment plans; However, according to Danuser, the differences that distinguish artificial intelligence in general cannot be interpreted in terms of specific cellular characteristics, making artificial intelligence difficult to use in clinical practice.
To meet this challenge, Danuser and his team used artificial intelligence to look for differences in the images of high and low metastatic potential melanoma cells, and then reengineered the artificial intelligence results to find out which image features were responsible for the differences .
Using tumor samples from seven patients and available information about their disease progression, including metastasis, the researchers filmed a video of approximately 12,000 cells randomly living in Petri dishes, which generated approximately 1,700,000 raw images. The research team then used an artificial intelligence algorithm to find 56 different abstract numerical features of the images.
The researchers found a feature that precisely differentiated between cells with high and low metastatic potential: by manipulating the abstract numerical feature, the researchers created artificial images that exaggerated the visible features of the metastasis that are invisible to the human eye.
The highly metastatic cells produced slightly more pseudopod enlargements and had more light scatter, an effect that could be due to subtle rearrangements of cell organelles.
“To further demonstrate the usefulness of this tool, the researchers first classified the metastatic potential of human melanoma cells that were frozen and grown in Petri dishes for 30 years and then implanted in mice. that spread easily between animals, while those predicted to have little metastatic potential spread little or no, ”the press release said.
Danuser said this method requires further investigation before it can be incorporated into clinical care; However, Danuser added that it might be possible to use AI to differentiate between important features of cancer and other diseases.
Source link
