Machine Learning to diagnose Deep Vein Thrombosis in Primary Care

Research statement
Using a handheld Ultrasound device connected to an Artificial Intelligence Software to enable non- specialist clinicians to diagnose DVTs in the community setting.
I would like to propose to organise and delivery the current deep vein thrombosis (DVT) diagnostic pathway into the primary care setting by implementing diagnostics through artificial intelligence (AI) guided point of care ultrasound (POCUS). The current diagnostic pathway and delivery of care for DVT patients is laborious, time-consuming and costly. With the new AI technology, hospital presentations could be avoided, as well as the usual multiple appointments with specialists. Patients could receive faster a diagnosis and treatment if DVT diagnostics were performed exclusively in primary care. An AI technology which guides the operator through a DVT scan can enable non-specialist operators, such as GPs or community nurses, to diagnose or rule out DVT at the point of care.
Those typically affected by DVT are elderly, immobile, multi-morbid patients, often isolated in their homes. It is especially difficult and distressing for this patient group to follow the current diagnostic pathway from their GP to A&E and from there to the specialist sonographer. Diagnosis is often delayed, causing increased complications such as Pulmonary Embolism (PE) and Post Thrombotic Syndrome (PTS). Due to the delay, GPs are often forced to prescribe unnecessary anticoagulation treatment, which lays the ground for an increased risk of bleeding.
A newly developed AI technology proved to be a reliable diagnostic tool in my MSc study. Implemented as an alternative to the current diagnostic pathway, it could take a significant burden away from already strained hospitals by enabling DVT diagnosis and treatment to be executed in primary care. The current laborious pathway could be replaced in many cases by a simple pathway, undertaken exclusively in primary care. This would reduce the workload of specialists, freeing them up for other tasks. Faster diagnostics would also reduce complications such as PE and PTS, which often require ongoing care and treatment. Furthermore, the financial weight of DVT diagnostics on our NHS could be remarkably reduced.
In my recent MSc project ‘Benefits of Machine Learning to diagnose Deep Vein Thrombosis compared to gold standard Ultrasound’, I was able to prove the AI device ‘AutoDVT’ is reliable diagnostic tool for DVT. A summary of my MSc thesis is pending to be published in the ‘Advanced Journal of Vascular Medicine’. In my PhD project I would like to investigate to what extent the AutoDVT could be implemented in primary care and to measure time and cost improvements.
Already in my BSc project, I highlighted that our healthcare system is overstretched by an insufficient number of physicians, an ageing society with more complex problems and a steadily increasing number of patients. I pointed out that we need to act now and open new health professions, such as physician associates, to ease the growing pressure on the healthcare system. The urgent need for improvement has been exacerbated by Brexit, which has led to a reduction in health care specialists in the UK. Furthermore, illnesses in the population have been left untreated due to the recent Covid pandemic, when many people avoided hospitals and clinics for fear of contracting Covid-19. As a result, people who could have received quick and cheap treatment in the early stages of an illness could now need more extensive and costly treatment.
Shifting DVT diagnostics to primary care settings ensures the well-being and safety of DVT patients and could take a huge financial burden off the NHS.