Font size: 18px14px
Print
AI Model Revolutionises Cancer Diagnosis and Treatment
Context:
Scientists at Harvard Medical School have introduced CHIEF (Clinical Histopathology Imaging Evaluation Foundation), a groundbreaking AI model set to transform cancer diagnosis and treatment.
More on News:
- Current AI systems are generally designed for specific tasks, like detecting cancer or predicting a tumour’s genetic profile, and are often limited to a few cancer types.
- CHIEF operates with flexibility similar to large language models like ChatGPT, covering a broad range of diagnostic tasks such as cancer detection, prognosis prediction, and treatment response assessment.
- Tested on 19 cancer types, CHIEF demonstrates the potential to surpass existing AI tools in various aspects.
Training and Performance of CHIEF
- The model was first trained on 15 million unlabeled images, focusing on sections of interest to recognise key features, followed by further training on 60,000 whole-slide images covering a wide range of tissues, including:
- Cancer Types: Lung, breast, prostate, colorectal, stomach, esophageal, kidney, brain, liver, thyroid, pancreatic, cervical, uterine, ovarian, testicular, skin, soft tissue, adrenal gland, and bladder.
- CHIEF outperformed current state-of-the-art AI models by up to 36% in several key areas.
- Including cancer cell detection, tumour origin identification, predicting patient outcomes, and identifying gene and DNA patterns related to treatment response.
- It demonstrated remarkable versatility and accuracy, performing effectively regardless of whether tumour cells were obtained via biopsy or surgical excision, or how they were digitised.
- This adaptability enables CHIEF to be applied across various clinical settings, overcoming the limitations of current models.
Key Highlights:
- CHIEF achieved nearly 94% accuracy in detecting cancer across 15 datasets with 11 different cancer types.
-
- It reached an impressive 96% accuracy in detecting cancers from biopsy data, including esophageal, stomach, colon, and prostate cancers.
- For surgical specimens, CHIEF maintained over 90% accuracy in detecting cancers from slides of tumours in the colon, lung, breast, endometrium, and cervix.
-
- Traditional DNA sequencing for tumour samples is costly and time-consuming, CHIEF offers a faster, cost-effective alternative by identifying cellular patterns suggestive of specific genomic aberrations directly from microscopic images.
- It effectively predicted patient survival from tumour histopathology images taken at initial diagnosis, distinguishing between longer-term and shorter-term survival across various cancer types and patient groups.
Implications:
- CHIEF’s ability to detect cancer and predict molecular profiles with high accuracy represents a significant advancement in AI-driven cancer diagnostics.
- By offering a rapid and precise alternative to traditional genomic profiling, CHIEF could improve treatment strategies and patient outcomes across a variety of cancer types.
Future Directions:
- Expanding training to include images from rare diseases and non-cancerous conditions.
- Incorporating pre-malignant tissue samples to enhance early detection.
- Integrating more molecular data to improve the model’s ability to identify varying cancer aggressiveness.
- Developing predictive features for novel cancer treatments and their potential side effects.
AI can help in tackling Disease ‘X’
- Artificial Intelligence (AI) is proving to be a crucial tool in addressing infectious diseases and potential future outbreaks, such as “Disease X,” an unknown pathogen that could trigger a global epidemic similar to COVID-19.
- The Canadian AI company BlueDot exemplifies this by accurately predicting the spread of COVID-19 by analysing data from news reports and animal health information.
- This early warning system enabled health officials to prepare for the outbreak before it became widespread.
- AI enhances the fight against Disease X by improving outbreak prediction, accelerating pathogen detection, and advancing drug and vaccine development, while also optimising healthcare responses through better forecasting and resource management.
- However, challenges such as data quality, bias, and privacy concerns need to be addressed to fully leverage AI’s potential in managing infectious diseases.
