Jan Lukas Robertus

My Topol fellowship problem / project:

Every patient’s case is subject to multiple decisions as part of their clinical pathway. These decisions involve a combination of weighing up evidence by expert clinicians and the application of national guidelines, with multiple clinical decisions being made before the patient undergoes diagnosis and treatment.

An example of one such decision point is the Multi-disciplinary Team meeting (MDT). MDTs depend on participating consultants, other medical team members, available patient data such as imaging, up-to-date investigations and the subsequent interpretation of national pathway guidelines.

Healthcare resources are increasingly limited – more so as we deal with the intense pressure of the current COVID-19 pandemic. Artificial intelligence (AI) can be used to support patients and clinicians by strengthening the decision-making that is part of diagnosis and treatment for each patient and by improving time efficiency and increasing uniformity of decisions between different institutes that deal with similar patient groups.

An AI-assisted clinical decision support system identifies information ‘signals’ that can be used to minimise uncertainty in complex, non-linear disease management.

Current ‘unassisted’ decision points such as MDTs and the connected clinical patient pathways before and after the MDTs, are not equipped to capture expertise in a knowledge framework that is dynamic and flexible in a uniform way while simultaneously learning within a framework of multiple hospitals and institutes.

I strongly believe that our healthcare systems will fundamentally improve with the implementation of digital healthcare. Three years ago, I developed and introduced a concept project using AI-assisted clinical pathways at the Royal Brompton and Harefield Hospitals, setting up a collaboration with IBM with support from Royal Brompton and Harefield Clinical Group, Guy’s and St Thomas’ NHS Foundation Trust.
Together with support from the Royal Brompton and Harefield Hospitals, we have now created a dedicated team of clinicians and together with IBM our team is developing and testing this system for patients and clinicians in four pathways: Inherited Cardiac Conditions, Pulmonary Hypertension, Lung Cancer and Trans-Aortic Valve Implantation.

We see these four pathways as proof of concepts and we aim to expand our system to all other patient care pathways in the near future.

Advantages for the AI assisted MDT process:

1. Adaptive decision-support platform that incorporates AI-functionality to automatically extract relevant clinical information from Electronic Healthcare Records (EHR) and facilitates rapid clinical decisions and strong guideline-compliant treatment approaches.
   2) Ensure each patient and their clinical team can focus available time and resources and have faster access to required documentation and case summaries available for diagnosis and treatment when needed.
   3) Flagging, tracking and coordinating the clinical needs, diagnosis and treatment of patients who are in multiple different pathways and in cases of inherited disease, the families of patients who may enter a clinical pathway separately.
   3) Introduce medical staff hospital wide to applied AI in healthcare and specifically in the NHS and large scale healthcare system.

Through the use of an AI assistant for verifying data completeness and continues automated analysing of areas of improvement, the overall process synchronises and draws from clinical and healthcare data to strengthen guideline-based decision support, bridging the inherent problems across the NHS, stemming from a lack of health data harmonisation.

Issues related to scalability within healthcare systems:

1. Application on a national level within the NHS or other large healthcare systems, joining together in a network that learns from different outcomes and provides accessible data and metrics for facilitating evaluation of process and outcome.
2. Identifying and creating opportunities for development of AI tools in the direct interpretation of clinical data and imaging data (histology and imaging modalities). Such applications will be used for prediction of treatment efficiency and disease progression leading to more efficient planning of treatment for each patient.
3. Delivery of the same personalized medicine to patients through digitized experience called from a library of knowledge across health data sets, providing a digital platform that will strongly support building a health data space and improve the quality and acceptance of AI-generated evidence in decision making in research and healthcare delivery.

The project is structured in multiple phases, each of which will produce a result that will be valuable in its own right, and that will provide a basis for the subsequent phases.

I will use the time protected by the TOPOL fellowship to further develop and manage the phases of the project.