Article of the Month – March 2023
March 1, 2023FDA Updates
March 1, 2023PathAI
A medical technology company with a technical pathology program capable of helping with research, drug development, global health solutions, and treatments.
PathAI is a medical technology company with a technical pathology program capable of providing faster and more accurate diagnoses. Furthermore, the software is currently being developed to help with research, drug development, global health solutions, and treatments. PathAI’s AI has two primary goals: to assist pathologists in making more reliable diagnoses and to predict better how patients will respond to a pathologist’s prescribed therapy based on patient tissue characteristics.
PathAI Principles
Clinical Precision: Empowering pathologists to make more accurate, standardized diagnoses using consistent, AI-driven methodology.
Drug Development: Accelerating drug development with pharmaceutical companies by increasing analytical capabilities and automating certain tasks so that scientists can focus on drugs’ more strategic biological implications.
Global Health: Applying the technology in global health. With a project funded by the Gates Foundation, PathAI is developing AI-driven applications that can provide pathologic diagnoses to developing nations at a low cost.
As PathAI grows, its data gravity has become an issue. PathAI searched for an industry-leading partner that could help reduce challenges associated with data gravity, such as complexity and cost while providing greater flexibility and a more efficient solution for secure data movement and processing. Despite Their background as a cloud-native company, they realized they could benefit from a hybrid cloud infrastructure.
PathAI selected Digital Realty due to the company’s scalable and flexible design options, unparalleled footprint in the data-dense Northern Virginia metropolitan area, availability of renewable energy solutions, and ability to deploy high-density, AI-ready infrastructure rapidly.
The different technologies are involved in the PathAI platform, from homemade tools to the usual suspects, Keras and TensorFlow. They have automated and scaled this process end-to-end using orchestration tools like Kubernetes, which provides a rich interface to the customers using Python, Django, and Flask. On the front end, PathAI has a modern and modular Vue.js framework with a well-developed pattern library to make interface engineering powerful and reliable. Overall, they have excellent tools throughout to ensure reliability, scale, privacy, security, and engineering sanity
PathAI recently presented ML models that predict breast cancer biopsies’ homologous recombination deficiency status (HRD). AI-powered ML models were trained to identify HRD directly from hematoxylin and eosin-stained breast cancer tumor biopsy slides from the cancer genome atlas (TCGA) and performed with high accuracy. The Human Interpretable Features-based model was trained using thousands of expert pathologist annotations of cell- and tissue-level features of the TCGA images to predict HRD status from HIF-based correlations. In contrast, the end-to-end model learned to predict HRD status directly from biopsy images.(1)
PathAI’s Financial Aspects
PathAI announced that it has secured strategic investments from the Merck Global Health Innovation Fund (Merck GHIF) and Bristol-Myers Squibb. These investments complete PathAI’s Series B investment round. The investment follows and extends PathAI’s series B investment, which was led by venture capital firms General Atlantic and General Catalyst, with strategic participation from LabCorp, a leading global life sciences company that provides comprehensive clinical laboratory and end-to-end drug development services.
PathAI will use the funding to bolster and accelerate its growing clinical development capabilities. These investments serve to deepen PathAI’s relationships with these two leaders in the development of cancer therapies.(2)
Liver Pathologies Treatment and Disease Progression Assessed by AI Models
PathAI uses machine learning (ML) and deep learning techniques to predict histological features of liver disease severity and is committed to enhancing the accuracy of diagnosis and the efficacy of other diseases.
One of the most important studies included HBV-infected patients with chronic HBV infection without cirrhosis regression following long-term tenofovir disoproxil fumarate (TDF) therapy where ML-based pathology revealed histologic features consistent with underlying NASH (Nonalcoholic Steatohepatitis).
Another study exhibited that ML models used in patients with PSC (Primary sclerosing cholangitis) accurately identified significant fibrosis inflammation.(3)
AI Digital Pathology Reveals Changes Related to the Pathogenesis and Progression of Nonalcoholic Steatohepatitis (NASH) and Chronic Hepatitis B Virus Infection (CHB)
ML can assess liver disease severity and identify features associated with disease progression. They can also estimate the hepatic venous pressure gradient (HVPG) in patients with NASH (nonalcoholic steatohepatitis) related cirrhosis and detect fibrosis from digitized slides. Important studies showed that ML models identified histological features suggestive of underlying fatty liver disease in patients with chronic Hepatitis B (CHB) and high serum alanine aminotransferase (ALT) despite virologic suppression.(4)
ML Models Reveal Treatment-Induced Changes in the NSCLC Tumor Microenvironment in Samples from LCMC3 Study
PathAI recently announced that their ML model was applied on NSCLC samples from the LCMC3 trial by Genentech to identify prognostic biomarkers in the tumor microenvironment and perform an AI-directed pathologic response assessment.
This study enrolled 181 participants with resectable, untreated stage IB to select IIIB NSCLC to evaluate the response to atezolizumab as a neoadjuvant treatment. Biopsies were collected from all patients before starting treatment, and surgical resections were collected after treatment from 159 patients. A significant pathologic response (less than 10% of viable tumor cells present) was achieved in 21% of subjects, and a complete pathologic response (0% of tumor cells were present after treatment) was observed in 7% of eligible subjects. PathAI’s AI analysis and tissue quantification will be presented at a future meeting.(5)
