What’s Going on in Digital Pathology

Getting ahead with ChatGPT
 

As artificial intelligence (AI) continues to evolve, researchers at the Dana-Farber Cancer Institute and Weill Cornell Medicine explore the possibilities of utilizing ChatGPT in digital pathology research (PMID: 38987117). Their study showcases the abilities of ChatGPT in assisting pathologists with complex software to analyze tissue samples, enhancing disease diagnosis and guiding treatment. However, the authors conclude that the full potential of domain-specific text-generative AI tools is still unfolding, commenting,  “We will soon see how these systems allow researchers to increase effectiveness of scientific investigation.”

Crystal clear
 

The clarity and magnification of microscopes has been limited by optical lens quality for hundreds of years. Previous research resulted in fourier ptychographic microscopy (FPM) to increase space-bandwidth, but the reconstructive methods are prone to failure under excessive abbreviations. Now, researchers at the California Institute of Technology showcase a new computational microscopy technique that removes guesswork and provides true high-resolution images (PMID: 38830852). The Angular Ptychographic Imaging with Closed-form (APIC) method has all the advantages of FPM without its weaknesses – allowing for clear images over a large field of view. 

Multidimensional mapping
 

Existing 2D histological staining methods are unable to trace the origins of precancerous pancreatic lesions that can cause cancer. In response to this bottleneck, researchers at Johns Hopkins Medicine have developed a 3D genomic profiling technique for identifying pancreatic intraepithelial neoplasias (PanINs) (PMID: 38693266). This machine-learning pipeline provides quantitative 3D histological reconstruction at single-cell resolution, highlighting the genetic relationships between and within PanINs. Now that PanINs have been identified, providing a foundation for early detection and interception of pancreatic cancer, researchers plan to  target them and investigate context in other organ tissues.

Tools of the trade
 

Research confirms that “one size fits all” is not possible in treating mental health conditions. In response to a growing demand for personalized treatment, two grants of over $10 million are aiding the investigation of the use of a smartphone app, AI, and other digital assessment tools for diagnosis and treatment of depression (https://bit.ly/4f4EyEJ). With BiAffect, a smartphone app created by teams at UIC, researchers hope to assess patients between clinic visits to identify additional subtypes of depression and build a dataset for further digital analysis. By stratifying patients, researchers hope to build better treatments at individual circumstances.