In recent years, the rapid advancement of artificial intelligence (AI) has created a wave of excitement and anticipation within the healthcare industry. AI has the potential to revolutionize patient care, improve diagnostic and prognostic accuracy, enhance treatment planning, and streamline administrative tasks. However, for AI to truly deliver on its promises, it is essential that AI tools undergo rigorous clinical validation. As both industry and government leaders continue to determine the proper regulatory oversight in the use of AI in healthcare, it is often difficult to find the middle ground between embracing innovation and sticking to the “old way” because it is trusted. As the CEO of an AI company that is focused on personalized cancer care, I believe in creating balance through prioritizing clinical validation and I encourage the healthcare industry to do the same.
By ensuring any AI technology that is adopted is clinically validated, leaders can be confident that the new tools they are implementing will ensure patient safety, provide accuracy, have a clinical impact, and be adopted by clinicians.
Patient Safety and Reliability
Clinical validation allows for AI models to be subjected to in-depth testing against large and diverse datasets and within real-world settings. They can also have their performance reviewed against established benchmarks and gold standards to ensure accuracy. By testing the performance of AI algorithms within real-world clinical scenarios, researchers can identify potential pitfalls and biases that may impact patient outcomes. One example is in prostate cancer care, where a new AI-derived biomarker was developed using a unique algorithm that assesses digital images from a patient’s biopsy and learns from a patient’s clinical data. The AI combines this information to predict whether a patient will benefit from hormone therapy and can estimate long-term outcomes. Models such as this are developed and validated with data from thousands of patients and terabytes of digital images. This is particularly important for ensuring that AI models exhibit robust performance across diverse patient populations, accounting for all different types of demographic variations and comorbidities.
Historically, clinical trials have underrepresented minority populations, and there are long-standing issues of bias in the healthcare field. With clinical validation, this issue can be addressed head-on by proactively using diverse datasets. For example, prostate cancer disproportionately impacts African-American men, with this population making up 16.3% of prostate cancer patients, but only an average of 9.4% of clinical trial participants. Given this disparity, studies have been completed to validate AI-derived biomarkers where 20% of the dataset was from the African-American population, which represented real-world scenarios.
Accuracy
Through the training and validation process, AI has been further enabled to more accurately prognosticate outcomes when compared to utilizing current guidelines alone. For example, when prognosticating prostate cancer, clinicians gather as much information as possible and use guidelines to determine a treatment plan for their patient. There is limited room for personalized cancer care using this method. Through rigorous training and validation, AI can use each patient’s clinical information and pathology slides to provide greater insights into how the disease will progress and which patient may benefit from specific treatments. The AI model can outperform the recommendations of the current guidelines available for treating men with localized prostate cancer. It can also predict which patients are likely to benefit from hormone therapy. By providing more accurate and individualized results, AI can help both clinicians and patients make more confident treatment decisions.
Clinically Impactful
In order for an AI tool to truly have a clinical impact, it needs to improve both the lives of patients and clinicians. Rigorous validation studies involving real-world patient populations help evaluate the clinical utility and effectiveness of AI tools in diverse healthcare settings. This evaluation provides valuable insights into the benefits and limitations of AI integration, allowing healthcare clinicians to make more informed decisions. With these data, clinicians feel informed and confident with their treatment decisions.
Better Adoption by Clinicians
Some clinicians are hesitant about adopting new technologies into their workflow, and this can especially be true for AI, which many feel is too complex and difficult to understand. When a tool has been through clinical validation, clinicians are more likely to be interested in adopting the technology, because it has been through rigorous testing and has proven results. This is why clinical validation plays a crucial role in building trust and confidence in AI technologies. By emphasizing clinical validation, we can dispel the myth that AI is too complex to understand and highlight transparent and accountable solutions that align with established clinical guidelines and practices.
Clinicians also want to feel confident in the tools that they are using. Clinical validation creates evidence that the technology is effective, accurate and reliable. This confidence for both clinicians and patients allows for greater shared decision-making when deciding on a treatment plan, which may help enhance the patient experience and improve outcomes.
As a healthcare leader, the balance between embracing innovation and being a skeptic of this new wave of technology can be difficult to manage. The way to find this balance is through clinical validation. Through rigorous validation, we can ensure the safety, accuracy, and clinical impact of AI algorithms, enabling healthcare providers to make informed decisions and deliver improved patient care. By promoting clinical validation as a gold standard, we pave the way for the responsible and effective integration of AI in healthcare, advancing the quality and efficiency of care delivery.
About Andre Esteva
Andre Esteva is a researcher and entrepreneur in medical artificial intelligence. He currently serves as CEO of ArteraAI, and was previously Head of Medical AI at Salesforce Research. He has worked at Google Research, Sandia National Labs, GE Healthcare, and has founded four companies.
