Benefit of Machine Learning to Diagnose Deep Vein Thrombosis Compared to Gold Standard Ultrasound

The study coordinator aims to compare gold standard deep vein thrombosis (DVT) diagnostic performed by a specialist sonographer to a scan by a non-specialist with a newly developed an automated DVT (AutoDVT) detection software device.

The title of the project is: Benefit of Machine learning to diagnose Deep Vein thrombosis compared to gold standard Ultrasound.

Currently the process from the DVT symptom begin, to diagnosis and then treatment is all but not straightforward. It implements a laborious journey for the patient from their general practitioner (GP) to accident and emergency (A&E), then to a specialist sonographer.

However, handheld Ultrasound devices have recently become available and they have been implemented with a machine learning software. The startup company ThinkSono developed a software which is hoped to divide between thrombosis and no thrombosis. In this single-blinded pilot study, patients which present at St Mary's DVT Clinic will be scanned by the specialist and then by a non-specialist with the machine learning supported device. The accuracy and sensitivity of this device will be compared to the gold standard.

This would mean that DVT could be diagnosed at point of care by a non-specialist such as a community nurse or nursing home nurse, for example beneficial for multimorbid confused nursing home patients. This technology could reduce A&E crowding and free up specialist sonographer to focus on other clinical tasks. These improvements could significantly reduce the financial burden for the National Health System (NHS).

The AutoDVT has a CE (as the logo CЄ, which means that the manufacturer or importer affirms the good's conformity with European health, safety, and environmental protection standards) Certificate under the directive 93/42/ European Economic Community (EEC) for medical devices. It is classified in Class 1 - Active Medical Device - Ultrasound Imaging System Application Software (40873).

Furthermore, following standards and technical specifications have been applied: British Standard (BS) European Norm (EN) International Organisation for Standardisation (ISO) 13485:2016, BS EN ISO 14971:2012, Data Coordination Board (DCB)0129:2018, ISO 15233-1:2016.

Study Overview

• Status: Withdrawn
• Conditions: Deep Vein Thrombosis; Deep Venous Thrombosis of Leg
• Intervention / Treatment: Device: 3-compression Ultrasound scan with AutoDVT software integrated Ultrasound probe

Detailed Description

"AutoDVT" is a software system designed to assist non-specialist operators, such as nurses, general practitioners (GP) and other allied health professionals in the diagnosis of DVT. The software utilises a "machine learning" algorithm as described below.

This study aims to improve the current laborious, time consuming and expensive diagnostic DVT pathway.

Venous thrombosis (VT) commonly occurs in the deep leg veins as well as the deep veins of the pelvis. DVTs can be divided into above knee (iliac, femoral, popliteal) and below knee (calf veins).

DVT is well recognised to cause globally significant morbidity and mortality both at the time of diagnosis and post-diagnosis. Between 30 - 50 percent of patients diagnosed with DVT will go on to develop a post-thrombotic syndrome, which has a significant impact on patients' long-term quality of life. Patients with DVT are also at risk to develop a fatal pulmonary embolism (PE). According to Charity Thrombosis United Kingdom (UK) dies every 37 seconds a person of a VT in developed countries.

Between 75-88 percent of suspected DVT cases, when fully investigated, are negative. The cost for diagnosing DVT over a decade ago was between 42-202 British Pound (£), such that the cost to the NHS of investigating all patients who present with DVT symptoms was approximately £175 million annually as stated in the study 'Non-invasive diagnosis of deep vein thrombosis from ultrasound imaging with machine learning' by Prof. Kainz from Imperial College London.

It is important to note that this value does not take into account any additional indirect costs such as time lost from work, hospitalisation, treatment costs and costs for repeat ultrasound scans. It is difficult to diagnose a DVT by clinical exam alone. The current standard approach to diagnose a proximal DVT involves an algorithm combining pre-test probability (Wells Score), D-dimer (blood) testing, and compression ultrasonography (typically a three-point compression examination).

There are new handheld ultrasound (US) probes available, meaning only the US probe is required for diagnostic purposes in conjunction with a mobile phone or tablet. At present, although the new handheld probes are smaller and are better suited for point of care diagnosis, they still require an experienced radiologist or sonographer to perform the three-point compression exam.

This means, that these devices can only be used wherever specialists such as radiographers or radiologists are based. However, due to recent advances in "machine learning", a software has now been developed for these 'app-based' probes that can assist non-specialist healthcare professionals to carry out the compression US exam with minimal training and divide between DVT and not DVT.

The previous data-collecting study for this device at Oxford University Hospital (OUH) was primarily used to improve the AutoDVT software but it also highlighted in a small pilot study that this technology had a similar diagnostic test accuracy to standard compression US. The study outlined in this protocol will test this hypothesis.

• Study Type: Observational

Contacts and Locations

Study Locations

• United Kingdom
  • London, United Kingdom, W2 1NY
    • Thrombosis Clinic, The Bays at St.Mary's Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years to 90 years (Adult, Older Adult)
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

• Sampling Method: Probability Sample

Study Population

Patients, which present at the DVT-clinic, The Bays at St.Marys Hospital with DVT symptomatic and requiring a scan

Description

Inclusion Criteria:

1. The participant has capacity to consent and consent is obtained
2. The participant is an adult (18 or older in the UK)
3. The participant has symptoms suggestive of a deep venous thrombosis
4. The diagnostic DVT algorithm indicates that an ultrasound is needed

Exclusion Criteria:

A patient will not be eligible for this study if they fulfil one or more of the following criteria:

1. Participant cannot consent
2. Participant is under age of 18
3. No data of D-dimer result
4. The participant is found to have a distal DVT during the US scan (retrospective exclusion)
5. Patient did not sign consent form

Study Plan

How is the study designed?

Design Details

• Observational Models: Other
• Time Perspectives: Cross-Sectional

Number of groups / cohorts

3

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Goldstandard DVT diagnostic through specialist sonographer; Patients with DVT symptomatic of lower limb which present at the DVT Clinic the Bays at St. Mary's Hospital will receive Goldstandard Scan through specialist.	Device: 3-compression Ultrasound scan with AutoDVT software integrated Ultrasound probe; The study coordinator, Miss Kerstin Saupe will perform a three-point compression ultrasound scan (USS) of the upper leg with the AutoDVT software. The AutoDVT software will store the results of the scan for retroperspective analysis and review. Effectiveness of AutoDVT as diagnostic tool will be evaluated in perspective to different patient groups, eg. patients with adipositas or maligner disease,
DVT diagnostic with Ultrasound probe with AutoDVT Software; Above patients will receive a second Ultrasound scan by a non-specialist with AutoDVT device.	Device: 3-compression Ultrasound scan with AutoDVT software integrated Ultrasound probe; The study coordinator, Miss Kerstin Saupe will perform a three-point compression ultrasound scan (USS) of the upper leg with the AutoDVT software. The AutoDVT software will store the results of the scan for retroperspective analysis and review. Effectiveness of AutoDVT as diagnostic tool will be evaluated in perspective to different patient groups, eg. patients with adipositas or maligner disease,
Retrospective data evaluation; result of AutoDVT is equivalent to diagnosis of specialist sonographer. Results specifically highlighted in patients with adipositas or malignant disease.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity of AutoDVT diagnostic software compared to diagnose by a trained sonographer	The compressibility of veins is measured in a 3-compression ultrasound performed with a n AutoDVT diagnostic software as tool. The results will be compared to the gold standard scan by a specialist.	Day 1, no follow up

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity of AutoDVT as diagnostic tool in obese patients	The compressibility of veins is measured in a 3-compression ultrasound performed with an AutoDVT diagnostic software as tool. The results will be compared to the gold standard scan by a specialist.	Day 1, no follow up
Sensitivity of AutoDVT as diagnostic tool in patients with previous history of malignancy	The compressibility of veins is measured in a 3-compression ultrasound performed with an AutoDVT diagnostic software as tool. The results will be compared to the gold standard scan by a specialist.	Day 1, no follow up

Collaborators and Investigators

• Sponsor: Imperial College London
• Investigators: Study Director: Mohammed F Aslam, MBA, PhD, Imperial College London

Publications and helpful links

General Publications

• Kainz B, Heinrich MP, Makropoulos A, Oppenheimer J, Mandegaran R, Sankar S, Deane C, Mischkewitz S, Al-Noor F, Rawdin AC, Ruttloff A, Stevenson MD, Klein-Weigel P, Curry N. Non-invasive diagnosis of deep vein thrombosis from ultrasound imaging with machine learning. NPJ Digit Med. 2021 Sep 15;4(1):137. doi: 10.1038/s41746-021-00503-7.

Helpful Links

• Thrombosis in UK in 2019

Study record dates

Study Major Dates

• Study Start (Actual): March 1, 2022
• Primary Completion (Actual): August 1, 2022
• Study Completion (Actual): August 1, 2022

Study Registration Dates

• First Submitted: February 16, 2022
• First Submitted That Met QC Criteria: March 18, 2022
• First Posted (Actual): March 21, 2022

Study Record Updates

• Last Update Posted (Estimate): January 30, 2023
• Last Update Submitted That Met QC Criteria: January 26, 2023
• Last Verified: January 1, 2023

More Information

Terms related to this study

• Keywords: machine learning; ultrasound; diagnostic; thrombosis
• Additional Relevant MeSH Terms: Cardiovascular Diseases; Vascular Diseases; Embolism and Thrombosis; Thrombosis; Venous Thrombosis
• Other Study ID Numbers: IRAS313321

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: No

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No