The post #InTheSpotlight | Beyond the hype, what might AI actually mean for healthcare in SA? appeared first on Spotlight.
BY Jesse Copelyn – IN SUMMARY:
- AI in Healthcare: Artificial intelligence (AI) is being tested and used in healthcare to assist with diagnostics, research, and even therapy, with varying levels of success. AI can perform tasks traditionally done by doctors, such as detecting tumors in x-rays and diagnosing illnesses.
- AI in Diagnostics: AI has shown potential in diagnosing diseases. For example, Google tested an AI model that provided more accurate diagnoses than doctors using the same patient information.
- AI as a Therapist: AI-powered Chatbots are being used for tasks like answering medical questions, booking appointments, and even providing therapy via apps. However, it’s unclear if AI therapy is truly effective due to limited studies.
- AI in Medical Research: AI is helping with medical research, such as Google DeepMind’s AlphaFold, which predicts protein structures. This has helped researchers working on vaccines like malaria.
- Challenges of AI in Healthcare: AI still faces challenges, such as generating incorrect information (“hallucinations”). The impact of AI in low-resource areas (like rural clinics) is also unclear.
- AI in TB Screening in South Africa: AI is making a big impact in screening for tuberculosis (TB) in South Africa, where mobile x-ray units use AI to detect abnormal lung images. AI-based tools called CAD (computer-aided detection) can identify TB and, or better than, human radiologists.
- AI Performance in TB Detection: In a 2021 study, the AI tool qXR captured 90% of TB cases at a specific threshold score, outperforming human radiologists. The tool analyzes x-ray images and assigns a score indicating the likelihood of TB.
- AI in Silicosis Screening: AI is also being used to screen for silicosis, a lung disease affecting miners. The National Department of Health is adopting CAD systems for both TB and silicosis to improve diagnostic accuracy and patient outcomes.
- Regulatory Challenges: Regulations for AI in medical devices are outdated. The South African Health Products Regulatory Authority (SAHPRA) hasn’t yet registered AI-based devices, and there’s concern that AI systems’ evolving nature complicates regulatory oversight.
- Data and AI Training: AI is sometimes trained on outdated data. For example, AI tools might miss subclinical TB (cases without symptoms) because they were trained on data from symptomatic patients.
- International Support: International agencies like Global Fund and USAID are funding AI-assisted x-ray machines for TB screening in South Africa, deployed in mobile clinics and hotspot areas.
- AI and Public Health Systems: South Africa’s health data systems are fragmented, which could hinder AI deployment. Building internal AI expertise and ensuring interoperability in digital systems will be critical for success in using AI.
- Vision for the Future: AI holds promise for improving South Africa’s healthcare, especially in tackling diseases like TB and silicosis. However, the transition will be challenging without proper infrastructure and technical expertise.
