Pramana Hematopathology AI

Using the latest scanning technology from Pramana, we are developing AI tools to improve diagnostic accuracy and workflow efficiency through the digitization of complex tissue specimens and the development of novel machine learning algorithms. This partnership focuses on one of the most challenging domains in anatomic pathology—bone marrow biopsy interpretation—for which traditional diagnostic methods are limited by variability, complexity, and throughput constraints.

Pramana’s SpectralHT scanners, equipped with edge-based AI computing, are deployed to digitize hematopathology slides at scale with volumetric fidelity and real-time quality control. These scanners generate high-resolution images suitable for algorithmic interrogation, while simultaneously streamlining slide processing through autonomous imaging and in-line analysis.

Leveraging ARUP’s domain expertise in hematopathology and robust digital infrastructure, the Research and Innovation (R&I) team leads the design, annotation, and development of AI models that assist with diagnostic interpretation. This includes curating and labeling large-scale, real-world datasets using expert-reviewed clinical cases, training deep learning models to recognize morphological patterns, and optimizing them for on-device execution. By integrating model inference directly into the scanner workflow, the team is enabling near real-time decision support at the point of image acquisition.

This collaboration illustrates the practical convergence of clinical-grade AI and digital pathology systems. It exemplifies how domain-informed algorithm development, combined with scalable edge deployment, can enhance diagnostic precision, reduce inter-observer variability, and drive efficiency in routine and specialized pathology workflows. The initiative also serves as a foundation for future commercialization, with models developed for seamless integration into laboratory environments in alignment with evolving regulatory pathways.

Through this work, ARUP and Pramana are redefining diagnostic workflows in hematopathology and setting new standards for AI-enabled clinical diagnostics.

Jalapheno

Jalapheno helps medical directors and variant scientists analyze whole genomes and exomes rapidly. Modern genome sequencing can identify millions of sequence variants in an individual, but finding which ones are medically relevant is a complex and time-consuming task. A single genome can sometimes require hours of manual review and research to accurately characterize the variants. The Jalapheno algorithm makes this task easier by identifying which genes in the genome are most likely to be associated with the clinical findings for a patient. Using these results, scientists can focus their efforts on areas of the genome most likely to be associated with the patient’s individual condition. By filtering down the millions of variants to just a few, Jalapheno helps patients receive accurate results faster. The algorithm works by combing public databases and journal articles to find information about clinical disease symptoms, known as “phenotypes,” and genes known to be associated with them. Jalapheno then assembles millions of these data points into a large graph of gene-phenotype associations, structured in a way to make searches lightning fast. Today, Jalapheno is used in clinical genome and exome workflows at ARUP, helping patients get accurate results, faster.

Aide

Aide is a cloud-based annotation and collaboration platform designed to streamline the curation and analysis of WSIs for research and diagnostic development. Built specifically for the unique demands of digital pathology, Aide enables distributed teams of pathologists, scientists, and engineers to collaboratively annotate, tag, and manage large-scale WSI datasets with efficiency and precision.

Aide supports project-based workflows, allowing users to create and share collections of WSIs across teams. Within these projects, users can generate structured annotations, apply customizable metadata tags, and search across slides using tag-based and contextual filters. The platform’s web-based interface allows seamless access across institutions, making it ideal for multi-site collaboration and remote expert review.

By centralizing annotation efforts in a secure, cloud-native environment, Aide addresses a critical bottleneck in AI development pipelines: the creation of high-quality, expert-labeled training data. The system is designed to maintain annotation fidelity, facilitate inter-observer consensus, and ensure metadata consistency—all essential components for the reproducibility and scalability of machine learning workflows.

Aide’s modular architecture also supports downstream integration with AI model training, data versioning, and digital pathology viewers, making it a cornerstone in the broader ecosystem of ARUP’s digital pathology infrastructure. Whether supporting internal research projects or external collaborations, Aide is accelerating the development of clinically relevant AI tools by enabling efficiency and turning high-resolution insights into complex tissue-based datasets.