Precision Pathology for Cancer and Beyond Gets Faster, Sharper with AI

A new artificial intelligence technology that reads medical images with exceptional clarity could free up time for time-pressed clinicians to focus on crucial areas of illness diagnosis and image interpretation. The iStar (Inferring Super-Resolution Tissue Architecture) technology was developed by researchers at the University of Pennsylvania’s Perelman School of Medicine, who believe it can help clinicians diagnose and cure tumors that would otherwise go unnoticed. The imaging approach provides both very precise views of individual cells and a broader glimpse at how people’s genes function, allowing doctors and researchers to spot cancer cells that would otherwise be nearly invisible. This technique can be used to detect whether safe margins were obtained during cancer surgeries and to offer automatic annotation for microscopic pictures, paving the path for molecular disease diagnosis at that level.

The strategy was described in a study published today in Nature Biotechnology by Daiwei “David” Zhang, PhD, a research associate, and Mingyao Li, PhD, a professor of Biostatistics and Digital Pathology.

According to Li, iStar can automatically detect critical anti-tumor immune formations known as “tertiary lymphoid structures,” the presence of which correlates with a patient’s likely survival and favorable response to immunotherapy, which is commonly used to treat cancer and requires high precision in patient selection. According to Li, this suggests that iStar could be a useful tool for evaluating which patients would gain the most from immunotherapy.

iStar was created as part of the field of spatial transcriptomics, a relatively new field that maps gene activity within the space of tissues. Li and her colleagues trained the Hierarchical Vision Transformer, a machine learning tool, on typical tissue images. It starts by segmenting photos into phases, starting small and checking for fine features, then rising up and “grasping broader tissue patterns,” according to Li. A network within iStar is driven by the AI system and uses the information from the Hierarchical Vision Transformer to anticipate gene activity, frequently at near-single-cell resolution.

“The power of iStar stems from its advanced techniques, which mirror, in reverse, how a pathologist would study a tissue sample,” he said. “Just as a pathologist identifies broader regions and then zooms in on detailed cellular structures, iStar can capture the overarching tissue structures and also focus on the minutiae in a tissue image.”

Li and her colleagues tested the tool’s efficacy on a variety of cancer tissues, including breast, prostate, kidney, and colorectal tumors, as well as healthy tissues. During these tests, iStar was able to detect tumors and cancer cells that were difficult to distinguish by eye. Clinicians may be able to detect and diagnose more difficult-to-see or difficult-to-identify tumors in the future, with iStar acting as a support layer.

Aside from the therapeutic potential given by the iStar technique, the tool moves at an extraordinarily fast rate when compared to other, similar AI technologies. For example, when employed with the team’s breast cancer dataset, iStar completed its analysis in just nine minutes. In comparison, the best competitor AI tool took more than 32 hours to perform a comparable analysis.

In other words, iStar was 213 times faster

“The implication is that iStar can be applied to a large number of samples, which is critical in large-scale biomedical studies,” says Li. “Its speed is also important for its current 3D and biobank sample prediction extensions.” A tissue block in 3D may consist of hundreds to thousands of serially sliced tissue slices. The speed of iStar allows us to reconstruct this massive amount of geographical data in a short period of time.”

The same is true with biobanks, which house hundreds, if not millions, of samples. This is where Li and her team intend to focus their future research and iStar expansion. They want to aid researchers in gaining a better knowledge of the microenvironments within tissues, which could lead to more data for diagnostic and therapy purposes in the future.

Source Link