Researchers develop novel AI algorithm for digital pathology analysis

Emerging field of digital pathology enhanced by new algorithm that up-levels classification.

Digital pathology is an emerging field which deals with mainly microscopy images that are derived from patient biopsies. Because of the high resolution, most of these whole slide images (WSI) have a large size, typically exceeding a gigabyte (Gb), and this means that typical image analysis methods cannot efficiently handle them.

Longevity.Technology: Digital pathology involves sharing and interpreting pathology digital information, including both slides and data. Glass slides are captured with a scanning device, and the resulting high-resolution images can be shared and viewed on a computer screen or smartphone leading to faster diagnoses and better-informed treatment, all feeding into Internet of Things.

However, with great resolution comes great file size, so, identifying a need for improvement, researchers from Boston University School of Medicine (BUSM) have developed a novel artificial intelligence (AI) algorithm based on a framework called representation learning to classify lung cancer subtype based on lung tissue images from resected tumours.

Analysing a WSI in patches works well for tumour identification, but if the goal is to identify the entire tumour region, capture the connectivity of the tumour microenvironment or characterise the stage of disease, then WSI-level information is needed.

“We are developing novel AI-based methods that can bring efficiency to assessing digital pathology data. Pathology practice is in the midst of a digital revolution,” explains corresponding author Vijaya B Kolachalama, PhD, FAHA, assistant professor of medicine and computer science at BUSM. “Computer-based methods are being developed to assist the expert pathologist. Also, in places where there is no expert, such methods and technologies can directly assist diagnosis [1].”

The researchers developed a graph-based vision transformer for digital pathology called Graph Transformer (GTP) that leverages a graph representation of pathology images and the computational efficiency of transformer architectures to perform analysis on the whole slide image [2].

“Translating the latest advances in computer science to digital pathology is not straightforward and there is a need to build AI methods that can exclusively tackle the problems in digital pathology”, explains co-corresponding author Jennifer Beane, PhD, associate professor of medicine at BUSM [1].

Using whole slide images and clinical data from three publicly available national cohorts, the researchers then developed a model that could distinguish between lung adenocarcinoma, lung squamous cell carcinoma, and adjacent non-cancerous tissue. Over a series of studies and sensitivity analyses, they showed that their GTP framework outperforms current state-of-the-art methods used for whole slide image classification.

This novel algorithm will enable more accurate and timely diagnoses, with WSIs sent worldwide to various pathologists. As well as diagnosing problems, this method could improve research experiments and speed up clinical trials. Machine learning is key in longevity as we gather more and more data about how aging happens, its causes and its consequences – being able to crunch and leverage that data is vital.

The research team also believes their machine learning framework has implications beyond digital pathology. “Researchers who are interested in the development of computer vision approaches for other real-world applications can also find our approach to be useful,” says Beane [1].


Photograph: Anna Tarazevich/Pexels