5 Ways to Use Generative AI in Healthcare

PwC predicts that healthcare spending will increase by 7% in 2024. This growth is primarily driven by healthcare worker burnout, labor shortages, disputes between payers and providers, and inflation. To ensure effective patient care without excessive operational costs, the industry is exploring innovative technologies such as generative AI in healthcare.

Accenture reports that 40% of healthcare providers' working time can be improved with AI, and a recent Forbes article suggests that this technology could save the U.S. healthcare sector at least $200 billion in annual costs.

Generative AI in healthcare uses machine learning algorithms to analyze unstructured data, such as patient medical records, medical images, audio recordings of consultations, etc., and create new content similar to what it was trained on.

In this article, our generative AI consulting firm will explain how this technology can help healthcare organizations.

Use Cases of Generative AI in Healthcare

1. Enhancing Medical Training and Simulation

2. Assisting in Clinical Diagnostics

3. Contributing to Drug Development

4. Automating Administrative Tasks

5. Generating Synthetic Medical Data

   
   

Visit our blog to learn more about how Gen AI can be used in business.

Enhancing Medical Training and Simulation

Generative AI in healthcare can create realistic simulations that replicate a wide range of health conditions, allowing medical students and professionals to practice in a safe, controlled environment. AI can generate models of patients with various diseases or assist in simulating surgical operations or other medical procedures.

Traditional training involves pre-programmed scenarios, which can be limiting. AI, on the other hand, can quickly generate patient cases and adapt in real-time, responding to decisions made by trainees. This creates a more complex and authentic learning experience.

Real-Life Example

• The University of Michigan has developed a generative AI model in healthcare that can create various scenarios for simulating sepsis treatment.

• The University of Pennsylvania used a generative AI model to simulate the spread of COVID-19 and test different interventions. This helped researchers assess the potential impact of social distancing and vaccination on the virus.

  
  

Assisting in Clinical Diagnostics

Here’s how generative AI in healthcare can aid in diagnostics:

• Creating High-Quality Medical Images: Hospitals can use generative AI tools to enhance the diagnostic capabilities of traditional AI. This technology can convert low-quality scans into high-resolution medical images with great detail, apply AI algorithms to detect anomalies, and present results to radiologists.

  
  

• Disease Diagnosis: Researchers can train generative AI models on medical images, lab tests, and other patient data to detect and diagnose early manifestations of various diseases. These algorithms can identify skin cancer, lung cancer, hidden fractures, early signs of Alzheimer's disease, diabetic retinopathy, and more. Additionally, AI models can identify biomarkers that may trigger specific disorders and predict disease progression.

  
  

• Answering Medical Questions: Diagnosticians can turn to generative AI in healthcare if they have questions instead of searching for answers in medical books. AI algorithms can process large volumes of data and quickly generate responses, saving valuable time for doctors.

  
  

Real-Life Examples

• A group of researchers experimented with Generative Adversarial Networks (GANs) to extract and enhance features in low-quality medical scans, converting them into high-resolution images. This approach was tested on brain MRI scans, dermoscopy, fundus imaging, and heart ultrasound images, showing superior accuracy in anomaly detection after image enhancement.

  
  

• Another example is Google’s Med-Palm 2 AI, trained on the MedQA dataset, achieving an 85% accuracy rate when answering relevant medical questions. Google acknowledges that the algorithm still needs improvement, but it’s a strong start for generative AI as a diagnostic assistant.

  
  

Contributing to Drug Development

According to the Congressional Budget Office, the process of developing new drugs costs an average of $1–2 billion, including failed drugs. Fortunately, there is evidence that AI can reduce the time required for drug development and screening by nearly half, saving the pharmaceutical industry around $26 billion in annual costs. Additionally, this technology can cut costs associated with clinical trials by $28 billion per year.

Pharmaceutical companies can use generative AI in healthcare to accelerate drug development by:

• Designing and creating new molecules with desired properties that researchers can later evaluate in laboratory settings.

• Predicting the properties of new drug candidates and proteins.

• Creating virtual compounds with high binding affinity to a target, which can be tested in computer simulations to reduce costs.

• Predicting the side effects of new drugs by analyzing their molecular structure.

For more information on the role of AI in drug development and how it supports clinical trials, visit our blog.

Real-Life Examples

• The rise of strategic partnerships between biotech companies and AI startups is an early sign that generative AI is taking over the pharmaceutical industry.

• Recently, Recursion Pharmaceuticals acquired two Canadian AI startups for $88 million. One of them, Valence, is known for its generative AI capabilities and will work on developing drug candidates based on small and noisy datasets, which are insufficient for traditional drug development methods.

• Another interesting example comes from the University of Toronto. A research group created the ProteinSGM generative AI system, which can generate new realistic proteins after studying images of existing protein structures. This tool can produce proteins at high speed, and another AI model, OmegaFold, is then used to evaluate the potential of the generated proteins. Researchers reported that most of the new generated sequences fold into real protein structures.

  
  

Automating Administrative Tasks

This is one of the most well-known use cases of generative AI in healthcare. Studies show that the burnout rate among U.S. physicians has reached a staggering 62%. Physicians suffering from this condition are more likely to be involved in incidents that endanger their patients and are more prone to alcohol abuse and suicidal thoughts.

Fortunately, generative AI in healthcare can partially relieve the burden on physicians by optimizing administrative tasks. It can simultaneously reduce costs associated with administration, which, according to HealthAffairs, accounts for 15%-30% of total healthcare spending. Here’s what generative AI can do:

• Extract data from patient medical records and populate the appropriate medical registries. Microsoft plans to integrate generative AI into the Epic EHR. This tool will perform various administrative tasks, such as responding to patient messages.

• Transcribe and summarize patient consultations, fill in the relevant EHR fields with this information, and create clinical documentation. Microsoft Nuance has integrated the generative AI technology GPT-4 into its clinical transcription software. Doctors can already test the beta version.

• Generate structured health reports by analyzing patient information such as medical history, lab results, scans, etc.

• Develop treatment recommendations.

• Answer physicians’ questions.

• Find optimal time slots for appointments based on patients’ needs and doctors’ availability.

• Create personalized appointment reminders and follow-up emails.

• Review medical insurance claims and predict which ones are likely to be denied.

• Create surveys to collect patient feedback on various procedures and visits, analyze them, and generate actionable insights to improve healthcare delivery.

  
  

Real-Life Example

• Navina, a healthcare AI startup, has created a generative AI assistant that helps doctors more effectively manage administrative duties. This tool can access patient data, including electronic health records, insurance claims, and scanned documents, provide status updates, recommend treatment options, and answer doctors' questions. It can even generate structured documents such as referral letters and progress notes.

• Navina has already raised $44 million in funding, indicating strong interest from the medical community.

  
  

Generating Synthetic Medical Data

Medical research relies on access to large volumes of data on various health conditions. Such data is critically lacking, especially when it comes to rare diseases. Additionally, collecting such data is expensive, and its use and dissemination are regulated by privacy laws.

Generative AI in healthcare can create synthetic data samples that can supplement real health datasets and are not subject to privacy regulations since the health data does not belong to specific individuals. AI can generate EHR data, scans, etc.

Real-Life Examples

• A group of German researchers created the GANerAid AI-based model for generating synthetic patient data for clinical trials. This model is based on the GAN approach and can produce medical data with desired properties, even if the original training dataset was limited in size.

  
  

• Another group of scientists experimented with generative AI to synthesize electronic medical records. The researchers were motivated by restrictive data privacy regulations and the inability to effectively share patient data between hospitals. They built the EHR-M-GAN model, which could output heterogeneous EHR data of mixed types (i.e., containing both continuous and discrete values) that realistically represent patient trajectories.

  
  

Ethical Considerations and Challenges of Generative AI in Healthcare

While tech and consulting giants continue to invest in AI, we also see renowned AI experts, including Tesla CEO Elon Musk and OpenAI CEO Sam Altman, warning about the risks associated with the technology. So, what challenges does generative AI bring to healthcare?

• Bias: The effectiveness of AI models is only as good as the dataset they were trained on. If the data does not adequately represent the target population, it will leave room for bias against underrepresented groups. Since generative AI tools are trained on vast volumes of patient medical histories, they inherit any biases present there, making it challenging to detect, let alone eradicate.

  
  

• Lack of Regulation: Despite AI raising significant ethical concerns, there are currently no official regulations governing the use of this technology. The U.S. and EU are working on formalizing relevant policies, but this will not happen in the near future.

  
  

• Accuracy Issues: AI does make mistakes, and in healthcare, the cost of such errors is quite high. For example, large language models (LLMs) can hallucinate. This means they can produce syntactically plausible outputs that are factually incorrect. Healthcare organizations will need to decide when to allow errors and when to demand that the AI model explains its conclusions. For example, if generative AI is used to assist in cancer diagnosis, doctors are unlikely to accept such a tool if it cannot justify its recommendations.

  
  

• Accountability: Who is responsible for the final health outcome? Is it the doctor, the AI provider, the AI developers, or some other party? The lack of accountability can negatively impact motivation and performance.

  
  

Ready to Enhance Your Medical Practice with Generative AI?

• Generative AI algorithms are becoming increasingly powerful. Robert Pearl, Clinical Professor at Stanford University School of Medicine, said:

  "ChatGPT doubles its power every six months to a year. In five years, it will be 30 times more powerful than it is today. In ten years, it will be 1,000 times more powerful. What exists today is like a toy. The next generation of tools is expected to have a trillion parameters, which, oddly enough, approximately matches the number of connections in the human brain."

  AI can be a powerful ally, but if misused, it can cause significant harm. Medical organizations should approach this technology with caution. If you are considering implementing AI-based solutions in healthcare, here are three tips to get started:

  
  

  Prepare Your Data: Even if you choose a pre-trained, ready-to-use AI model, you may still need to retrain it on your own dataset, which must be high-quality and representative of the target group. Always keep medical data secure and protect patient privacy. It would be helpful to disclose the dataset on which the algorithm was trained, as this helps understand where it will perform well and where it might fail.

  
  

  Take Control of Your AI Models: Foster a culture of responsible AI in your organization. Ensure that people know when and how to use the tools and who takes responsibility for the final outcome. Test generative AI models on use cases with limited impact before scaling them to more sensitive applications. As mentioned earlier, generative AI can make mistakes. Decide where a small percentage of errors is acceptable and where you cannot afford them. For example, 98% accuracy might be sufficient for administrative applications, but it is unacceptable in diagnostics and patient care. Develop a framework that governs the use of generative AI in healthcare within your hospital.

  
  

  Help Your Staff Embrace and Utilize the Technology: AI still needs human guidance, especially in the highly regulated healthcare sector. Human-in-the-loop remains a critical component for the technology's success. Medical and administrative staff will need to oversee AI models, so hospitals should focus on training people for this task. Staff, in turn, should be empowered to rethink their daily routines now that AI is part of them, using the freed-up time to create more value.

  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  

  #GenerativeAI #HealthcareAI #MedicalInnovation #AIinHealthcare #HealthTech #MedicalAI #HealthcareAutomation #AI #MachineLearning #DigitalHealth #PatientCare #HealthData #MedicalResearch #JustSoftLab