AI Summary: The article critiques the integration of artificial intelligence (AI) in healthcare, highlighting its tendency to reinforce existing power hierarchies and exacerbate vulnerabilities among marginalized populations. It argues that AI systems often emerge from profit-driven institutional contexts, leading to the encoding of social inequities and the potential for digital colonialism, particularly in Global South settings. The authors emphasize that health-related datasets are frequently incomplete and biased, resulting in underrepresentation of marginalized communities, which in turn affects the efficacy of AI models in clinical settings. This "data absenteeism" is framed as a form of epistemic violence, as it not only neglects but actively exploits these populations, undermining their health outcomes and perpetuating systemic inequalities.
National Taiwan University Hospital develops AI for pancreatic cancer metabolic profiling
AI Summary: Researchers from Academia Sinica and National Taiwan University Hospital have developed PanMETAI, an AI-integrated metabolomics screening platform aimed at early detection of pancreatic cancer. This diagnostic model utilizes a standardized liquid biopsy to extract and analyze approximately 260,000 metabolic signals from 500 microliters of blood serum, achieving high accuracy with area under the curve values of 99% for non-cancer and 93% for pancreatic cancer in validation studies. The model's approach, which captures global metabolic changes, enhances early risk identification and demonstrates reproducibility across diverse datasets. The development reflects over twenty years of clinical expertise and aims to improve early diagnosis not only for pancreatic cancer but potentially for other cancer types as well.