A Multi-center Cohort Study for Conventional Ultrasound Image Set Collection to Create a Data Set for Research Purposes.

A Multi-center Cohort Study for Conventional Ultrasound Image Set Collection to Create a Training Data Set for Research Purposes (Image Processing and Analysis, AI Model Training).

This study aims to collect and create a labelled ultrasound image data set containing ultrasound image series and video clips of patients that undergo routine ultrasound scans on lower limbs, because of suspected deep vein thrombosis.

The data will be used to train an AI model within ThrombUS+ project to achieve automated detection of deep vein thrombosis on conventional ultrasound scans.

Primary objectives:

1. Collect and curate imaging data from ultrasound scans of patients suspected for DVT.
2. Collect accompanying metadata on patient demographics, referral note, existing known medical conditions at the time of scan, diagnosis based on the scan, operator anonymized ID, metadata on the ultrasound equipment used.
3. Anonymize the data set according to established regulations to be used for research purposes and in specific for training an artificial intelligence model to achieve automated DVT detection.

Secondary objectives:

1. Describe the data set in the Argos/OpenAIRE tool and make it publicly available through the European Open Science Cloud (EOSC) portal via OpenAIRE, to be used by other researchers for image processing, analysis, and artificial intelligence (AI) model training.

Deep vein thrombosis (DVT) and its fatal complication pulmonary embolism (PE) afflict millions of people worldwide and are responsible for a large percentage of acute hospitalizations. Clinical assessment of DVT is notoriously unreliable because up to 2/3 of DVT episodes are clinically silent and patients are symptom free even when PE has developed. Symptomatic DVT events, that are eventually referred for ultrasound imaging are only the tip of the DVT iceberg. The subgroup of events that evolve to develop clinical indications cannot be accurately predicted and often lead to sudden death from PE regarded as "the leading cause of preventable death in hospitalized patients" and "the number one priority for improving patient safety in hospitals".

Deep vein thrombosis (DVT) is the formation of a blood clot within the deep veins, most commonly those of the lower limbs, causing obstruction of blood flow. In 50% of people with DVT, the clot is at some point detached from the vein wall and travels to the lung to cause pulmonary embolism. About 25% of people experiencing pulmonary embolism (PE) will die from it, making it the 3rd leading cause of cardiovascular death worldwide after stroke and heart attack. Even in patients who do not get PE, recurrent thrombosis and "post-thrombotic syndrome" are major causes of mortality and reduced quality of life.

Recent European population studies report DVT incidence of 70-140 cases/100,000 person-year, which translates to roughly 522,000 to 1.04 million cases per year in Europe. Respectively, Center for Disease Control and Prevention (CDC) reports around 900,000 DVT incidents per year in USA, with an estimate of 60,000-100,000 related deaths per year. Venous thromboembolism (that collectively defines DVT and/or PE) during hospitalization is the leading cause of disability-adjusted life-years (DALYs) lost in low- and middle-income countries, and the second most common cause in high-income countries, causing loss of more DALYs than nosocomial pneumonia, catheter-related bloodstream infections, and adverse drug events. No identifiable provoking risk factor is reported in about 25%-40% of DVT and pulmonary embolism incidents. Surgery is reported to account for 15% of the cases and especially orthopaedic surgery with postoperative rates of around 1% reported despite pharmacological thromboprophylaxis; immobilization is reported to account for 15% and cancer for about 20% of cases.

Early diagnosis of DVT is crucial and has been proven to prevent life-threatening complications (pulmonary embolism), minimize the risk of long-term disability (post-thrombotic syndrome, recurrent DVT), improve treatment outcomes, and reduce healthcare costs. Despite the progress made in ultrasound imaging and plethysmography techniques, there is a need for new methods to enable continuous monitoring DVT diagnosis in hospitalized and other high-risk patients at the point of care.

ThrombUS+ EU Horizon project brings together an interdisciplinary team of industrial, technology, regulatory, social science and clinical trial experts to develop a novel wearable device for operator free, continuous monitoring in patients with high DVT risk. The devices and software to be developed during this project are expected to achieve automated early DVT detection, provide a continuous assessment of DVT risk and support DVT prevention via extended reality and serious gaming. ThrombUS+ wearable is intended for use by postoperative patients in the ward, during long surgical operations, cancer patients or otherwise bedridden patients at home or in care units, and women during pregnancy and postpartum. ThrombUS+ will use big data sets for artificial intelligence (AI) training collected in the project via 3 large scale clinical studies and will validate the outcome in the clinical setting via 1 early feasibility study and 1 multi-center clinical trial.

This study (ThrombUS_Study_A) aims to collect and create a labelled ultrasound image data set containing ultrasound images series of patients that undergo routine ultrasound scans on lower limbs, because of suspected deep vein thrombosis.

The data will be used to train AI models within ThrombUS+ project to achieve automated detection of deep vein thrombosis on conventional ultrasound scans.

The data set will include negative scans, positive for deep vein thrombosis scans, positive for other diagnosis scans and scans of insufficient quality to aid towards diagnosis, together with imaging metadata and a set of labels for each scan. In addition, the dataset will include pseudonymized patient demographics, referral note, existing known medical conditions at the time of scan, diagnosis based on the scan, operator anonymized ID, and metadata on the ultrasound equipment and scanning protocol parameters.

The data set will be completely anonymized to be used for research purposes, in compliance with the General Data Protection Regulation (GDPR) and the European Health Data Space (EHDS) and the upcoming Artificial Intelligence Act (AIA). Furthermore, the anonymized data set will be described in the Argos/OpenAIRE tool and will be made available through the European Open Science Cloud (EOSC) portal via OpenAIRE, to be used by other researchers for image processing, analysis, and AI model training. This is a non-interventional diagnostic image data collection study.

Inclusion Criteria:

• Age ≥18 years.
• The participant has the capacity to consent, and consent is obtained prior to any study-specific procedures.
• The conventional diagnostic DVT algorithm indicates that an ultrasound is needed, or the patient has been referred for a scan on suspicion of DVT.

Exclusion Criteria:

• Patients with a known condition or reason that may potentially result in interrupting or stopping the ultrasound examination before its completion.
• Patients considered by their treating physician or the ultrasound operator as non-suitable for a standard ultrasound scan.
• Patients who have not signed the informed consent.