GST Exemption Announced for Cancer Treatment Gene Therapy

AIIMS Develops Affordable CAR-T Cell Therapy for Multiple Myeloma

Sangamo Therapeutics and Astellas Collaborate on Innovative Gene Therapy for Neurological Diseases

FDA Clears IND for Arbor Biotechnologies’ ABO-101 Gene Editing Therapeutic for Primary Hyperoxaluria Type 1

Fortis Healthcare Acquires Stake in Agilus Diagnostics

FDA Approves BRAFTOVI® Combination Regimen as First-Line Treatment for BRAF V600E-Mutant Metastatic Colorectal Cancer

The Role of Big Data in Precision Medicine

Caris Life Sciences Celebrates 1,000+ Precision Medicine Studies, Advancing Cancer Research and Patient Outcomes

OncoHelix and Qualisure Diagnostics Partner to Bring Thyroid GuidePx® Test to Canadian Patients

Gates Foundation Invests $50 Million in Tessera Therapeutics to Advance Gene Therapy for Sickle Cell Disease

FDA Approves Ensartinib (Ensacove) as First-Line Treatment for ALK-Positive NSCLC

Guardant Health and Boehringer Ingelheim Collaborate to Advance Liquid Biopsy CDx for HER2-Targeted Therapy in NSCLC

Cartherics Receives $300K G-Rex Grant to Advance Cancer Cell Therapy Manufacturing

Roche Revolutionizes Mass Spectrometry Diagnostics with Launch of cobas® Mass Spec Solution

Complete Genomics Partners with Eurofins Clinical Enterprise to Expand Sequencing Service Offerings

Home
/
insight
/
details

Pathology in the Age of Generative AI

NextEdge Admin

26 Jul 2024

2 min 31 sec

Precise and swift pathology is vital for detecting, diagnosing, and treating various conditions such as cancer and infectious diseases. Computational pathology has advanced significantly with AI, digital pathology slides, and extensive image datasets. In The Lancet Digital Health, Omar et al. discuss large language models (LLMs) like OpenAI’s ChatGPT-4, noting their potential in medical research but emphasizing the need for domain-specific AI tools. They introduce the Digital Pathology Assistant, which uses the PathML library to process large-scale pathology images. Lu et al. describe PathChat, a multimodal generative AI that integrates a vision encoder, multimodal projector, and LLM, excelling in diagnostic tasks across diverse tissue types. PathChat’s training code is publicly accessible for academic use. Li et al. validate AI applications with the Cpath TIL-score biomarker, linking high TIL density in ductal carcinoma in situ to increased recurrence risk and radiotherapy benefit.

Despite the promise of AI in digital pathology, challenges like model interpretability, workflow integration, and data bias need addressing. Under-representation of minority groups in training datasets is a significant concern, affecting AI tool quality and generalizability. This issue is critical for low- and middle-income countries (LMICs), where cancer fatalities are projected to rise significantly. Strategies to support digital pathology in LMICs include open datasets, open-source models, public-private partnerships, and long-term funding. The Lancet Digital Health advocates for advancements in AI tools, addressing biases, ensuring minority representation, and promoting global implementation of these technologies for improved healthcare outcomes.

Reference:
Omar et al., Lu et al., Li et al., The Lancet Digital Health