Benefit of machine learning to diagnose deep vein thrombosis compared to the gold standard ultrasound
| ISRCTN | ISRCTN24147434 |
|---|---|
| DOI | https://doi.org/10.1186/ISRCTN24147434 |
| EudraCT/CTIS number | Nil Known |
| Secondary identifying numbers | V1.0 |
- Submission date
- 11/05/2023
- Registration date
- 15/05/2023
- Last edited
- 02/06/2023
- Recruitment status
- No longer recruiting
- Overall study status
- Completed
- Condition category
- Circulatory System
Plain English Summary
Background and study aims
Deep vein thrombosis (DVT) is a term that describes blood clots (thrombi) that can form in the deep veins. The deep leg veins are commonly affected (such as the proximal veins: the femoral vein or the popliteal vein) or the deep veins of the pelvis. The standard approach to making a diagnosis involves an algorithm combining pre-test probability, a blood test called the D-dimer test, and the patient undergoing an ultrasound of the leg veins. Ultrasound is currently completed by a trained expert (e.g. sonographer or radiologist). However, handheld ultrasound probes have recently become available and they have enabled ‘app-based’ ultrasonography to be performed. ThinkSono has developed software (AutoDVT software) allowing non-specialists to perform DVT ultrasound, hoping it has the same accuracy for diagnosing DVT as the standard ultrasound. If this study has a positive outcome, it would mean that DVT could be diagnosed at the point of care by non-experts such as nurses, junior doctors, GPs and other healthcare staff. By diagnosing DVT early in the clinical pathway (for example, at GP practices), the technology could reduce emergency department admissions and free up specialists to focus on other clinical tasks. These improvements could also potentially reduce the financial burden of the DVT diagnostic service on healthcare systems.
Who can participate?
Patients aged 18 years and over, coming for a check to see if they have a DVT and have symptoms suggesting that they need an ultrasound scan
What does the study involve?
Participants undergo two compression ultrasound scans. One is carried out by a non-radiology staff member (e.g. a nurse) using AI software to guide them and another ultrasound scan will be carried out as already scheduled by a sonographer or radiologist.
What are the possible benefits and risks of participating?
This study will not benefit participants directly in the short term but it may benefit patients having an ultrasound for a DVT in the future. The results from this study will improve knowledge of how software may be able to help diagnose blood clots accurately and quickly. Ultrasound is a very safe method of confirming a DVT or not and is used already as standard care in hospitals. There are no risks of taking part. The scan does involve some pressing on the leg but if it is painful or participants want to stop they can let the researchers know.
Where is the study run from?
Helios Klinikum Emil von Behring, Klinik für Angiologie (Germany)
When is the study starting and how long is it expected to run for?
January 2022 to September 2022
Who is funding the study?
1. Imperial College London (UK)
2. Helios Klinikum Emil von Behring, Klinik für Angiologie (Germany)
Who is the main contact?
Sven Mischkewitz (Sponsor contact), hello@thinksono.com
Contact information
Scientific
University of Bristol
Bristol Medical School
Room 1.10 (Desk 3)
Canynge Hall
Clifton
Bristol
BS8 2PN
United Kingdom
 ORCID ID |
0009-0004-5767-8346 |
|---|---|
| Phone | +4915141406431 |
| kerstin.nothnagel@bristol.ac.uk |
Study information
| Study design | Non-randomized prospective study |
|---|---|
| Primary study design | Observational |
| Secondary study design | Cohort study |
| Study setting(s) | Hospital |
| Study type | Diagnostic |
| Participant information sheet | Not available in web format, please use contact details to request participant information sheet |
| Scientific title | Benefit of machine learning to diagnose deep vein thrombosis compared to the gold standard ultrasound |
| Study hypothesis | Measure the sensitivity and specificity of clinical decision-making based on images recorded by a non-specialist using AutoDVT AI-based software versus the ground truth produced by the local imaging specialist. |
| Ethics approval(s) | Approved on 15/03/2022, Ethics Committee of the Berlin Medical Association (Ethik-Kommission der Ärztekammer Berlin), (Working Committee Research II [Arbeitsausschuss Forschung II] Ärztekammer Berlin KdöR, Friedrichstr 16, 10969 Berlin, Germany; +49 30 408 06 26 01; ek@aekb.de), ref: Eth-07/21 |
| Condition | Proximal deep vein thrombosis |
| Intervention | Patient recruitment Patients will consent when a DVT ultrasound exam is requested in the angiology department. An AI-assisted scan with the AutoDVT software is performed by a non-specialist (nurse). A follow-up gold-standard scan is performed by a local specialist (compression ultrasound), i.e. angiologist. That same-day follow-up scan represents the standard of care. The images collected by the non-specialist are presented to a remote, qualified clinician who evaluated image quality according to the quality scale of the American College of Emergency Physicians (ACEP) and consequently, if the image quality is sufficient, assesses whether the veins of the patient are compressible, incompressible or indeterminate. Study arms Single-arm study. Every patient will receive an AI-guided scan which was followed up by a reference scan which represents the standard of care. Intervention provider The nursing staff carried out the AI-guided ultrasound scan. They had no prior ultrasound experience at all. The remote, qualified clinician assessing the images collected by the AI-guided scan are qualified to diagnose DVT, i.e., radiologists. Modes of delivery The AI-guided scan is performed face-to-face. The images that are presented to the remote, qualified clinician will be evaluated retrospectively via an internet platform. This remote, qualified clinician will not see the patient. The AI-guided scan was carried out in the rooms of the angiology department. |
| Intervention type | Device |
| Pharmaceutical study type(s) | Not Applicable |
| Phase | Not Applicable |
| Drug / device / biological / vaccine name(s) | AutoDVT |
| Primary outcome measure | Sensitivity and specificity measured using the AI-guided ultrasound and a local imaging specialist performing the gold-standard ultrasound exam at one timepoint. The gold-standard exam is performed on the same day. |
| Secondary outcome measures | Image quality of the AI-guided ultrasound assessed by remote qualified clinicians, according to the American College of Emergency Physicians (ACEP) scoring scale from 1 to 5, at one timepoint |
| Overall study start date | 01/01/2022 |
| Overall study end date | 01/09/2022 |
Eligibility
| Participant type(s) | Patient |
|---|---|
| Age group | Adult |
| Lower age limit | 18 Years |
| Sex | Both |
| Target number of participants | Minimum of 60 |
| Total final enrolment | 91 |
| Participant inclusion criteria | 1. Aged 18 years old and over 2. Subject can consent and consent has been signed 3. Subject has symptoms of DVT and ultrasound is indicated |
| Participant exclusion criteria | 1. Below the age of 18 years old 2. No Wells score was calculated prior to the ultrasound 3. Distal DVT 4. Did not consent |
| Recruitment start date | 18/04/2022 |
| Recruitment end date | 09/08/2022 |
Locations
Countries of recruitment
- Germany
Study participating centre
14165 Berlin Germany
Berlin
14165
Germany
Sponsor information
Industry
August-Bebel-Straße 88
Potsdam
14482
Germany
| Phone | +491754724848 |
|---|---|
| hello@thinksono.com | |
| Website | https://thinksono.com/ |
Funders
Funder type
Hospital/treatment centre
No information available
Results and Publications
| Intention to publish date | 01/09/2023 |
|---|---|
| Individual participant data (IPD) Intention to share | Yes |
| IPD sharing plan summary | Available on request |
| Publication and dissemination plan | Planned publication in a high-impact peer-reviewed journal, potentially in combination with other study data. |
| IPD sharing plan | The datasets generated and analysed during the current study will be available upon request from Sven Mischkewitz (hello@thinksono.com) if a request is made by relevant authorities and a statement about the use of the data has been made. Anonymised ultrasound data may be shared. No patient data will be shared due to patient confidentiality. |
Editorial Notes
02/06/2023: The following changes were made to the study record:
1. The study design was changed from 'Non-randomized prospective double-blind study' to 'Non-randomized prospective study'.
2. The primary study design was changed from Interventional to Observational.
3. The secondary study design was changed from Non-randomised study to Cohort study.
13/05/2023: Trial's existence confirmed by Ethics Committee of the Berlin Medical Association (Germany).
