Personalization has become a widely embraced strategy in industries such as marketing, retail, and healthcare, which gives rise to the new term "personalized medicine." Also known as precision medicine, personalized medicine is the customization of medical interventions, predictions, and preventive measures tailored to the unique characteristics of an individual. Notably, the incorporation of AI and machine learning (ML) helps healthcare professionals overcome the current challenges in personalized medicine.
Supporting clinical decisions with AI-driven data analysis
Healthcare professionals usually utilize Clinical Decision Support Systems (CDSS) to access evidence-based information to improve patient care. However, managing and interpreting extensive datasets becomes more complex without AI-driven solutions. As personalized medicine relies heavily on various patient data sets (including genetics, molecular profiles, and medical histories), organizing and retrieving such data can be resource-intensive and less streamlined. Additionally, manual analysis of intricate datasets introduces significant time and labour costs, potentially resulting in delays in treatment decisions.
Thus, AI-powered CDSS can quickly analyze vast patient data to provide personalized recommendations, improve diagnostic accuracy, and enhance treatment planning. When a healthcare professional inputs patient information into the system, AI-driven CDSS can quickly analyze patients’ genetic information and laboratory results. Instead of making decisions based on current generalized criteria of age and gender, AI and ML enable patients to be screened based on their unique health profiles. This can enhance the precision of medical interventions and optimize the timing and dosage of medication for each patient.
An example of healthcare professionals using AI-powered CDSS is xDECIDE, developed by oncology-focused AI platform company xCures, which combines human expertise with AI software to offer personalized treatment options for cancer patients. In recent years, providing precise cancer care has become more challenging due to an 11% increase in the number of genes and a 35% increase in variants that can influence treatment decisions, which makes the process more complex. The system utilizes real-world evidence, expert knowledge, and scientific publications to identify potential treatments tailored to each patient. xDECIDE employs natural language processing (NLP) and ML to analyze medical records and generate a ranked list of treatment options. In the final stage, human experts review the output to provide personalized treatment recommendations for patients and ensure accuracy.
Early diagnosis and disease prevention with AI
One significant drawback of not utilizing AI in personalized medicine is the reduced predictive power for patient outcomes and treatment responses. Therefore, AI can tailor therapies based on patient’s characteristics, as AI algorithms are proficient in accurately predicting individual responses to treatment based on complex biological factors. By examining factors such as an individual's lifestyle and environmental context, healthcare providers can utilize AI to identify potential risk factors and suggest lifestyle adjustments to prevent the onset of certain diseases.
In 2021, Lark Health collaborated with Roche Diagnostics to establish an AI-powered lifestyle change program called the Lark Heart Health program. As heart disease is the leading cause of death in the US (accounting for approximately 1 in every 5 deaths), the Lark Heart Health program aims to provide patients personalized coaching virtually to prevent and manage critical risk factors for atherosclerotic cardiovascular disease (ASCVD) and coronary artery disease (CAD). Then, the program provided participants with customized AI-powered care coaching and educational lessons to promote better heart health.
After participating in the program for 40 days, the members experienced notable improvements in various aspects of self-confidence related to managing their heart health. Specifically, there was a 17% increase in confidence about knowing when to seek medical attention for heart disease, indicating a greater awareness of symptoms and warning signs. Despite being in its pilot stage, the program demonstrates how AI contributes to disease prevention with a personalized approach.
Ensuring safety, speed, and success for drug trials
Drug trials are essential in the field of personalized medicine as they are the bridge between scientific breakthroughs and real-world applications. As drug trials involve manual segmentation of patients based on genetic, molecular, and clinical factors, leveraging AI to identify patient groups can expedite drug trials and offer potential cost savings. AI can swiftly identify patients with the highest likelihood of responding positively to a new treatment, accelerating this development phase. Without AI, drug development can be slower and more costly. In fact, the average cost of bringing a new drug to market is USD 2.6 billion; AI can filter a group of patients for those most likely to respond positively to a new treatment, which can speed up this stage of development.
Furthermore, creating data models of individual patients, or "digital twins," allows researchers to conduct preliminary trials before embarking on an expensive one involving humans. This means researchers can assess the efficacy and safety of a drug in a simulated environment. Conducting preliminary trials using AI-generated digital twins can significantly reduce the time required to bring a drug to market. According to industry group PhRMA, it takes 10-15 years on average to develop one new medicine from initial discovery through regulatory approval. This streamlined process can accelerate drug development for personalized medicine and increase the success rate of drug trials in the real world.
Navigating ongoing challenges of AI in personalized medicine
Using health data in building datasets usually reveals inherent biases from limited diversity in sampling, missing values, and data collection methodologies. Consequently, when AI models are trained on such data, they might exacerbate these biases, potentially resulting in unfavourable decisions that affect certain demographic groups based on age, gender, race, geography, or economic status. Thus, some healthcare institutions are developing initiatives to alleviate this issue. For instance, the National Institutes of Health's All of Us program strives to recruit participants from diverse backgrounds to foster a more representative healthcare database. Furthermore, tools such as IBM's AI Fairness 360 toolkit can provide comprehensive metrics and algorithms to assist researchers in identifying and mitigating biases in datasets and models.
In addition to existing biases in health data, socio-environmental factors significantly influence the performance and effectiveness of AI models in clinical settings. For instance, a study conducted by Google Health evaluating an AI system for diabetic retinopathy screening encountered challenges when implementing the system in real-world clinical environments, particularly in clinics in Thailand. Different clinic conditions and workflows adversely impacted image quality and processing speed, while unstable internet connections and travel expenses for participants further complicated the deployment of AI systems. Hence, healthcare providers should validate AI models in different clinical environments and implement iterative feedback loops to enhance AI performance.
Data privacy is another concern within the healthcare industry, especially when reliance on AI technologies continues to grow. Therefore, ensuring a secure data storage, management, and sharing ecosystem is imperative to safeguard sensitive information. This requires adopting advanced technologies, collaborative efforts, and establishing robust regulatory frameworks and business models. Strategies such as encryption methods, strict access controls, and compliance with data protection laws (such as GDPR and HIPAA) are crucial in ensuring data's ethical and responsible utilization in precision medicine and healthcare AI applications.
Harnessing AI for personalized medicine
AI can be a helpful companion for healthcare professionals in personalized healthcare with safer drug trials, faster drug development, and more customized healthcare for patients with different genetic makeup and medical histories. As AI continues to evolve, the healthcare industry must comply with strict data guidelines and ensure human intervention to overcome its limitations.
