Application of Artificial Intelligence & Machine Learning in Clinical Research (Part-1)

Lot of us working in Clinical Research industry use Statistics. Very few of us use Artificial Intelligence (AI) and Machine Learning (ML). Therefore, it is important to understand difference between Statistics and AI. In nonprofessional terms, we can describe Statistics as something hard for humans and easy for computers. AI as something hard for computers but easy for humans. For Ex: humans struggle to manually calculate the p-value for large volume of data whereas machines can run the algorithm in fraction of minutes. Machines struggle with speech and image recognition, which humans can do easily. ML combines both AI and Statistics to tackle things hard for both humans as well as machines.

           Now if we look more closely at Statistical Modelling then we will realise that it involves several assumptions to evaluate whether active drug is better than Placebo or not. If we increase the number of input variables by millions where relationship between variables is not known and ask to determine the relationship with output variables then Classical Statistical Models will not work. It becomes cumbersome to evaluate the statistical significance of each input variable. Classical Statistics involves assumptions around data distributions before models are run.

            Machine Learning involves building Artificial Neural Networks, which can work in similar way as Human Neural Network works. Instead of making all decisions based on programmed data; ML is expected to make decisions based on learnings of the data fed in the past. ML starts with Supervised Learning, then moves onto Unsupervised Learning and finally to Reinforcement Learning.

           We have started seeing lot of attributes of Virtual Trials in existing Clinical trials due to COVID-19. There is increase in remote monitoring, increase in telemedicine, reduced site visits for patients. Going forward we will see more and more virtual trials as they will help in patient recruitment as well as retention. If we fast forward few years from now then for all clinical trials which do not involve invasive procedures; we might start seeing site less trials where data is reported via wearables, sensors, eCOA, Health Apps. We are also seeing regulators slowly adopting the new technologies and coming with some guidelines so that there is less ambiguity in the industry when it comes to adoption of new technologies. So the day is not far when we will see increased adoption of AI and ML in Clinical Trials. So this is the right time to get ourselves and our teams ready for the future of Clinical Trials by upskilling ourselves and keeping an open mind for the change. I would recommend you to follow Sanjeev Nair's series of Future Trends on YouTube as some of this text is adopted from his speech.