Harvard Researchers Develop ChatGPT-like AI Model “CHIEF” Capable of Detecting and Diagnosing Multiple Types of Cancer
Fiona Nanna, ForeMedia News
5 minutes read. Updated 11:11PM GMT Sun, 8 September, 2024
A groundbreaking artificial intelligence (AI) model, akin to ChatGPT but tailored for medical purposes, has shown remarkable potential in diagnosing and evaluating multiple types of cancer, according to recent findings. Researchers suggest that this new model could outperform existing deep learning methods, offering broader applicability in cancer detection and improving patient outcomes.
The model, known as the Clinical Histopathology Imaging Evaluation Foundation (CHIEF), has demonstrated a 36 per cent higher effectiveness in cancer detection compared to other AI-based approaches. It was designed to help clinicians diagnose a variety of cancers, determine the origin of tumors, and predict the likely course of patient outcomes.
Harvard Medical School researchers spearheaded the development of CHIEF, aiming for a more generalizable AI model that can tackle multiple diagnostic tasks. Unlike current deep learning models, which are often trained to handle specific diagnostic functions, CHIEF was created to work across a broader spectrum of cancer-related tasks.
How the CHIEF Model Works
Kun-Hsing Yu, an assistant professor of biomedical informatics at Harvard Medical School and senior author of the study, stated in an email to Euronews Health that the AI tool offers real-time second opinions for cancer diagnoses. “Our AI tool provides clinicians with accurate, real-time second opinions by considering a broad spectrum of cancer types and variations,” said Yu.
The model was trained on more than 15 million pathology images, enabling it to become adept at diagnosing cancers that may have unusual characteristics. The development process also involved refining the model with over 60,000 high-resolution images of tissue slides, which helped enhance its accuracy in genetic and clinical predictions.
CHIEF was tested on more than 19,400 images sourced from 24 hospitals and patient groups across the globe. The results, published in the scientific journal Nature, revealed that the AI model could analyze digital slides of tumour tissues and predict their molecular profiles based on visual features. It also demonstrated the ability to identify tumour traits that could inform treatment decisions.
The model achieved 94 per cent accuracy in detecting cancer across 11 different cancer types. In more specific cases, such as identifying colon cancer cells or predicting genetic mutations, CHIEF’s accuracy soared to 99.43 per cent.
A Leap Forward for AI in Oncology
The medical community views CHIEF as a significant advancement in applying AI to cancer diagnosis. Ajit Goenka, a professor of radiology at the Mayo Clinic in the US, who was not involved in the study, praised the model’s capabilities in an email to Euronews Health. “CHIEF could streamline preliminary diagnostic evaluations and provide pathologists with a valuable tool to highlight critical areas for further investigation,” Goenka explained.
However, Goenka also highlighted the importance of testing the AI model across diverse clinical settings to ensure its robustness. He cautioned that there might be biases in CHIEF’s performance due to its training on large, potentially non-representative datasets.
Next Steps: Regulatory Approval and Validation
Before CHIEF can be widely used in hospitals and clinics, the model must undergo regulatory approval and further testing in real-world environments. Yu and his team are already planning a prospective clinical study to validate CHIEF’s performance in clinical settings and ensure its reliability across various patient demographics and conditions.
Goenka reiterated the necessity of validating the model in practical, everyday medical practice: “Extensive validation in diverse clinical settings is crucial to confirm that CHIEF’s superior theoretical performance translates into tangible benefits for patient care.”
Looking ahead, the research team is also working to expand CHIEF’s capabilities to detect rare cancers, with the hope that this tool will revolutionize the way clinicians diagnose and evaluate tumors, ultimately improving treatment outcomes.
Meta Description:
Discover how a new AI model, CHIEF, designed by Harvard Medical School researchers, could revolutionize cancer detection by accurately diagnosing multiple types of cancer. Learn about its performance and future in clinical settings.
