The new deep learning model outperforms biomedical imaging analysis in understanding disease progression, pediatrics researchers at Children's Hospital of Philadelphia say.
By expediting the first step of spatial omics data analysis, the new artificial intelligence model provides detailed insights into how a disease develops and progresses at the cellular level and can advance precise diagnostics and targeted treatments, Children's Hospital of Philadelphia said in its announcement.
CHOP's open-source artificial intelligence software is available in a public repository for noncommercial use.
WHY IT MATTERS
The pediatric researchers developed a deep learning-enhanced biomedical imaging model, called CelloType, to expedite the identification and classification of cells in tissue images and then tested the biomedical imaging AI across a broad set of complex diseases, including cancer and chronic kidney disease.
CelloType is programmed to improve accuracy in cell detection, segmentation and classification, CHOP said, and is efficient for handling large-scale tasks like natural language processing and image analysis.
While CHOP's model requires training for segmentation and classification tasks, it can learn patterns and make predictions or classifications faster than previous approaches.
The researchers compared CelloType's performance against models that segment multiplexed tissue images, including Mesmer and Cellpose2, and detailed their results of the National Institutes of Cancer-funded research in Nature Methods.
"Unlike the traditional two-stage approach of segmentation followed by classification, CelloType adopts a multitask learning strategy that integrates these tasks, simultaneously enhancing the performance of both," they said in their report.
Certain cell types are either large or of irregular shape, presenting challenges to conventional segmentation methods. CelloType, which leverages transformer-based deep learning and automates the analysis of high-dimensional data, better captures complex relationships and context in tissue samples, they said.
CelloType uses AI to precisely outline objects in an image.
Kai Tan, the study's lead author and a professor in the Department of Pediatrics at CHOP, said in a statement that the "approach could redefine how we understand complex tissues at the cellular level, paving the way for transformative breakthroughs in healthcare."
THE LARGER TREND
There is a pressing need in spatial omics, a field that combines genomics, transcriptomics or proteomics with spatial information to map where different molecules are located within cells in complex tissues, for more sophisticated computational tools for data analysis, according to Tan.
Recent advancements have led to the analysis of intact tissues at the cellular level, allowing for unparalleled insights into the link between cellular architecture and functionality of various tissues and organs.
Using AI to improve the understanding of biomedical images can not only help clinicians treat patients but may also enhance patient access to advanced imaging and even predict diseases like cancer, thus health systems are embracing AI imaging tools.
While researchers in Norway and Denmark are using mammography images in national breast cancer screening programs to help predict diagnoses, Stamford Health's Heart & Vascular Institute announced in October that its patients will automatically receive coronary artery disease during any non-contrast chest CT scan and when their future risk indicators are elevated.
"This tool enhances our ability to detect early signs of cardiovascular disease and ensures that patients receive the follow-up care they need to prevent serious health outcomes," Dr. David Hsi, chief of cardiology and the institute's co-director, said in a statement.
One chief medical officer and pediatrics professor said that he believes that armed with AI and machine learning, healthcare providers can turn the tide for patients fighting complex diseases.
"Personalized genetic and epigenetic information can help tailor many medications to specific patients and to specific diseases. All of these omics involve huge amounts of data that information technology now can analyze in exquisite detail that can be assessed functionally through artificial intelligence and machine learning-derived algorithms," Dr. William Hay Jr., chief medical officer at Astarte Medical, a precision medicine company, told Healthcare IT News last year.
ON THE RECORD
"We are just beginning to unlock the potential of this technology," Tan said in a statement.