Sandeep Krishnan

An avid researcher in the field of medicine and healthcare, with an expertise in applying information technology in the area of critical real-life applications. Having worked for many years on the technology side for the implementation and maintenance of a hospital management system (MS4), including patient registration, eligibility and electronic claims, medical records, clinicals, payments et al, I possess a close working knowledge of the healthcare domain, including HIPAA, HL7 standards. My goal is to continue in this field through continued research and make contributions that can help in improving the life expectancy, wellbeing and quality of life globally. With a Post Graduate Diploma in Data Science, as part of my Thesis for the Masters, I am currently in researching for improving upon the existing works to improve the uncertainty estimates and prediction intervals for time-series forecasting, with a specific focus on its application in the area of clinical diagnostics. Especially in the medical domain, where the application of machine learning and artificial intelligence are now pervasive and have made significant contributions in improving the patient life, it is all the more important that the predictions that are being provided through the deep learning architectures deployed, are being accompanied by the relative uncertainty around the predictions so that sufficient risk assessment is done on that basis before a decision is made and also the relevant mitigations and contingencies are planned for the possible consequences of that decision. For example, when multiple differential diagnoses are noted based on the patient symptoms and lab work, it is imperative that any predictions obtained through leveraging artificial intelligence to identify one of the diagnoses as the primary and move to implementing the treatment protocol be sufficiently risk-assessed and validated with the uncertainty that exists around the identification itself. The provision and availability of sufficient information around the predictions would be essential for the medical practitioners to be able to make right treatment decisions as the stakes can be high where any treatment protocol undertaken based on erroneous/uncertain predictions can prove to be fatal. For this research, the data being leveraged shall be the MIMIC-III (Multiparameter Intelligent Monitoring in Intensive Care, v1.4) as the same has been leveraged for the earlier works (such as Alaa and van der Schaar, 2020), which will be extended as part of this study. The goal of the study is to forecast the daily observations of the white blood cell counts, which speaks to the condition and severity of the disease, while improving the predictive capability of the model through the control of false discovery rate and thereby improving upon the error rate among the significant results.