ICTworks

Generative artificial intelligence is poised to transform healthcare by revolutionizing clinical decision-making and improving patient outcomes.

This technology shows promise in enhancing patient care and advancing diagnosis and treatment options. GenAI could significantly improve healthcare through automation, better clinical decisions, and broader access to expertise. It offers the potential to make healthcare more efficient, equitable, and effective.

Subscribe Now for more digital health insights!

However, its role in healthcare remains poorly understood, with many challenges still unresolved and no clear path for its implementation in healthcare systems.

That is why I researched 6 different global health challenges and identified 10 examples of GenAI use cases.

These leading examples serve as a wonderful reminder that we need innovators willing to take measured risks and engage in careful planning to achieve technological advances that maximize benefits for patients, practitioners, and the health system as a whole.

Medical Education and Training

Generative AI enhances medical education by creating virtual patient cases based on diverse medical conditions, providing a comprehensive learning platform. This technology allows students to practice diagnoses and treatment plans in a risk-free environment, improving problem-solving skills.

GenAI can also create text, images, and even training program outlines, helping course designers create training faster and students learn medical topics quicker. Additionally, AI can track learner performance, offering feedback and helping educators refine their teaching strategies.

Here are two ways that humanitarian organizations are using GenAI in medical practitioner education and training programs.

• GenAI Producing Medical Training Data: Medical students in Rwanda have limited access to diverse clinical cases for training. The Digital Health Applied Leadership Program uses AI to generate a wide variety of virtual patient cases, exposing students to more complex and rare conditions. The blended learning approach and a focus on regional capacity building for digital health builds the country’s medical practitioners’ skills and capacities to handle a broader range of medical issues, enhancing patient care.
• GenAI Improving Medical Training Programs: Designers of medical training courses face challenges including how to develop high quality visual design assets, interactivity and videos, how to balance the demand and value of multilingual translation with its costs, and how to drive localization of learning materials. The TB DIAH project is utilizing GenAI in the instructional design process for an e-learning course focused on USAID’s TB performance-based monitoring and evaluation framework.LLMs for synthesis and translation have proven to be highly effective and cost saving. For example, ChatGPT can quickly and cheaply translate and synthesize participant feedback about course quality in three courses across four languages, and GenAI like Midjourney can create imagery for infectious diseases such as TB or HIV without stigmatizing human subjects.

Patient Education

Generative AI can create personalized educational content based on a patient’s condition, making complex medical concepts more accessible through visual aids and interactive learning experiences. AI can generate content at different reading levels, improving health literacy, and providing mental health support by addressing patients’ concerns. Additionally, AI-generated educational content can be provided in multiple languages, ensuring broader accessibility.

I identified at least three health organizations using GenAI chatbots to educate patients.

• GenAI Chatbot for HIV Prevention: It can be difficult to engage marginalized populations who are at the highest risk of becoming HIV positive. Amanda Selfie – a transgender chatbot – stimulated PrEP interest in adolescents and facilitated sensitive discussions on sex, STIs, and PrEP, and even identified individuals at higher risk for HIV in Brazil. It pioneered using artificial intelligence and online social platforms to create demand and access to health care services.
• GenAI for Women’s SRH: No one likes to talk about sexual and reproductive health with strangers, and in countries like India, women’s honest questions about SRH can be stigmatizing. Pinky Promise is a GenAI-powered chatbot clinic built by medical professionals. It also allows women to connect anonymously with each other. The app is making SRH information – be it on painful periods, fertility problems, abnormal discharge etc. – judgment-free and accessible.
• Chatbot to Understand Health Systems: Health systems are complex to navigate, and consumers, especially women, often disengage in frustration, leading to poor health outcomes at an individual and population level. askNivi is a conversation chatbot on messaging platforms like WhatsApp. Through a unique combination of behavioral science, real-time data segmented by demographic and behavioral attributes, and GenAI, askNivi empowers audiences to own their health outcomes, navigate health systems, and create better health outcomes for everyone.

Medical Diagnosis

Generative large language models (LLMs) can be trained on extensive medical records and imagery to identify disease patterns. LLMs are particularly useful in image reconstruction, synthesis, segmentation, registration, and classification, enabling the creation of synthetic medical images for training machine learning models.

LLMs reorganize information into user-friendly formats, improving patient comprehension. LLMs can also analyze electronic health records (EHRs), integrating data from multiple sources to provide a comprehensive view of a patient’s health, which is particularly useful in complex cases. While LLMs like OpenAI’s ChatGPT-4 demonstrate impressive medical knowledge, their outputs should complement, not replace, physicians’ clinical judgment.

Here are two examples of GenAI usage for medical diagnosis:

• AI Radiology to Diagnose Breast Cancer: Limited access to radiologists and diagnostic tools in remote areas like the Pacific Islands delays critical medical diagnoses. RAB uses AI to analyze mammogram images remotely, providing quick and accurate diagnostic support to local healthcare providers. This has significantly improved breast cancer diagnosis and treatment, providing high quality care to patients and supporting local clinicians in developing nations.
• Computer Vision to Identify Pathogens: It will take 400 years to address the shortfall of microbiologists in LMICs at the present rate of training. As a result, clinicians over-prescribe antibiotics, leading to antimicrobial resistance (AMR), a major threat to public health. MSF employs AI-powered diagnostic tools to assist local medical staff in identifying diseases and performing antimicrobial susceptibility testing. Antibiogo was validated to give high-quality diagnostic tests for 11 WHO priority pathogens, leading to rational antibiotic use and better patient care.

Personalized Medicine

Generative AI enables personalized medicine by analyzing a patient’s genetic makeup, lifestyle, and medical history to predict treatment responses. By identifying patterns in large datasets, AI can tailor treatment plans to individual needs, improving outcomes. AI can also be used in mental health, creating personalized cognitive behavioral therapy (CBT) scenarios that help patients manage their conditions.

Personalized medicine is especially important in developing countries since they often have large genetic diversity that is very different from the developed country populations where many drugs are designed and tested.

I found one example of GenAI used for personalized medicine, though there is great opportunity for it in the African context.

• Personalized Medicine for Cardiovascular Disease: Cardiovascular disease is a leading cause of death in Vietnam, yet personalized treatment plans are often unavailable. The AI4BetterHearts program – a global data collaborative on cardiovascular population health – uses AI to analyze patient data and recommend personalized treatment plans for cardiovascular diseases. This has led to more effective and tailored treatments, reducing mortality rates from heart disease in the region.
• Potential for Personalized Medicine in Africa: Recent research on personalized medicine advances in Africa, notes that African populations have huge genetic diversity and a significant need for individualized medicine to overcome African-specific health challenges. Researchers found only 620 papers on African patients, and only 21 that looked at clinical trials or randomized controlled trials for precision medicine. None of them expressly mentioned generative AI. In addition they only found 46 researchers from Africa affiliated with 34 African institutes.

Data Generation

Generative AI models are valuable for generating synthetic data that maintains patient privacy while providing realistic datasets for research and training. By simulating EHR data, these models address privacy concerns and improve machine learning models’ accuracy.

This approach is particularly beneficial in low- and middle-income countries when real-world data is scarce or access is restricted due to privacy issues. The ability to create diverse and representative datasets enhances machine learning models’ robustness, leading to new medical insights and discoveries.

Two examples that use GenAI to create synthetic data and improve health outcomes:

• Synthetic Data for Pandemic Modeling: During the COVID-19 pandemic, data privacy and scarcity in low-resource settings, limiting the effectiveness of AI-driven solutions for healthcare. UN Global Pulse used AI to create synthetic data that mimicked real patient data from a million Rohingya refugees living in temporary camps in Bangladesh, allowing researchers to develop and train models for pandemic response without compromising privacy. A visualization tool that presented the results of the epidemic modeling clearly to decision-makers enabled better understanding of the pandemic’s impact on vulnerable populations and supported more equitable healthcare solutions.
• Synthetic Data for Infectious Disease Modeling: Researchers used a deep generative model to generate synthetic data based on a small dataset – just 364 patients – in Vietnam. Then they used that synthetic data to build models that predict the onset of hospital-acquired infections based on minimal information collected at the patient ICU admission. The performance of the diagnostic model trained on the synthetic data outperformed models trained on the original and oversampled data using standardized prediction techniques, showing the promise of synthetic data, despite limitations in dataset size.

Drug Discovery

Generative AI accelerates drug discovery by generating new small molecules, nucleic acid sequences, and proteins with specific structures or functions. By analyzing the chemical structures of successful drugs and simulating variations, AI models can produce potential drug candidates more quickly than traditional methods.

This not only saves time and resources but also helps identify drugs that might be overlooked using conventional techniques. Moreover, AI can predict the efficacy and safety of new drugs by analyzing large datasets, reducing the time and cost of drug development. Additionally, AI can identify new biological targets for drug development, leading to more effective treatments.

Two examples of humanitarian uses for GenAI to discover new life-saving drugs:

• Discovering Drugs for Neglected Diseases: Traditional drug discovery methods for neglected diseases like sleeping sickness are slow and costly, leaving vulnerable populations without effective treatments. DNDi uses AI to repurpose existing drugs and discover new treatments faster by analyzing biological data and predicting drug efficacy. This approach has accelerated the development of new treatments for neglected diseases, improving patient outcomes in affected regions.
• Drug Discovery for Malaria Parasite: The high cost and time required for drug discovery limit access to effective treatments for diseases like malaria and tuberculosis in Africa. H3D leverages AI to identify new drug candidates for malaria prevention and treatment by screening large datasets of chemical compounds. This has led to research on compound MMV048, the first clinical anti-plasmodial candidate of a new class of compounds which target the malaria parasite PI4 kinase system.

Clinical Documentation

LLMs like ChatGPT-4 can automate the generation of patient data summaries, alleviating the documentation burden on healthcare providers and reducing burnout. These models can also analyze EHR data to produce concise summaries that highlight critical information, improving communication between healthcare providers and patients.

Automating documentation also reduces the risk of medical errors by ensuring important information is not overlooked. LLMs can also automate routine tasks like scheduling appointments and processing insurance claims, saving time and resources while improving patient experience.

Surprisingly, there are no examples of this activity in developing countries that I could find, although there is widespread use of GenAI tools for automated medical documentation in the USA. Automated medical documentation has already shown to reduce medical errors and reduce clinician burnout from doing what is perceived as administrative work.

Thanks to ChatGPT, Perplexity, Google Search, and GenAI healthcare researchers for helping me create this post.