Healthy volunteers for the collection of baseline ultrasound data and the establishment of valid reference and baseline values for the development of real time segmentation algorithms and automated report generation.
Conditions
Interventions
Group 1: Arm 1 name:
Healthy volunteers (data collection cohort)
Arm 1 description:
One time, non invasive ultrasound examination in healthy volunteers aged 18 years or older. A diagnostic abdominal
in newly included participants, a supplementary thyroid ultrasound may be performed only under separate consent. Ultrasound cross sectional images and short video sequences (cine loops) are captured a
data are subsequently stored in pseudonymized form.
Sponsors
Institut für Künstliche Intelligenz in der Medizin (IKIM)
Eligibility
Sex/Gender
All
Age
18 Years to No maximum
Inclusion criteria
Inclusion criteria: Healthy volunteers aged 18 years or older who are able to provide informed consent. Women of childbearing potential are eligible regardless of contraception use, as diagnostic ultrasound is non invasive and involves no radiation exposure or additional risk related to pregnancy status.
Exclusion criteria
Exclusion criteria: Age below 18 years or inability to provide informed consent. Participants with conditions or organ related abnormalities that would prevent standardized ultrasound acquisition of the target organs are excluded, as are patients.
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| The primary objective is the successful generation of a standardized, high-quality ultrasound dataset from 50 healthy volunteers to enable the development of Machine Learning algorithms for real-time applications. This encompasses several critical goals: 1. Successful Generation of a High-Quality Ultrasound Dataset: The successful creation of a high-quality dataset consisting of ultrasound cross-sectional images (DICOM) and short video sequences (CineLoops - DICOMS) and a video of the entire US examination (MP4 - US images only) from 50 healthy volunteers. 2. Minimum Data Volume Achievement: The collection of the estimated minimum data volume required for training the models, specified as at least 500 images per organ. This data collection must cover the targeted abdominal organs (kidneys, urinary tract, liver, biliary system, pancreas, spleen, and major abdominal vessels), with supplementary data collected for the thyroid gland in newly included participants (Per 1 december). 3. Development of real-time Segmentation and standardized automatic report generation algorithms: The development of real-time segmentation and standard report generation algorithms for the collected organs using the generated data and Machine Learning (ML). This is the core functional goal derived from the data collection. The study should result in multiple scientific articles, of which most are open source, e.g. automatic real-time segmentation per organ, foundational segmentation model, report generation, classification of cross-sectional images/CineLoops and even automatic classification of the cross-sectional images within the video. | — |
Secondary
| Measure | Time frame |
|---|---|
| 1. Dataset characteristics: Number and type of files per participant and organ (DICOM still images, DICOM cine loops, MP4 exam recordings) and total data volume. 2. Acquisition feasibility: Examination duration, time for export and pseudonymization and archival, plus frequency and reasons for incomplete acquisitions. 3. Protocol adherence and data quality: Completeness of required views per organ and descriptive quality ratings using the predefined checklist, including unusable data due to artifacts. 3. Optional thyroid component: Uptake (separate consent) and resulting data yield and quality for thyroid acquisitions (Starting first of december). 4. Annotation feasibility: Amount of annotated data, annotation time per organ or case, and inter annotator agreement on a subset if double annotation is performed. Exploratory algorithm outputs: Segmentation and classification metrics and report generation agreement on predefined evaluation subsets, plus real time feasibility metrics (for example latency or frames per second), reported descriptively. Incidental findings workflow: Number of incidental findings triggering the predefined documentation or referral pathway, reported descriptively if applicable. | — |
Countries
Germany
Contacts
Public ContactJan Egger
Institut für Künstliche Intelligenz in der Medizin (IKIM)
Outcome results
None listed