Dataset Structure

Schema Overview

The data types are selected to best suit a dataframe or SQL database for analysis.

Field name Recommended type Description Sample values
Gender Categorical / String The biological sex of the patient. Male, Female
Age String (Mixed) The age of the patient at the time of the study.
Note: Requires cleaning (see Data Fields).
025Y, 035Y
Modality String The imaging method used. All entries are US (Ultrasound). US
Description String Study type and anatomical region. Abdomen Complete, OB Ultrasound, Carotid Doppler
Size_raw String The file size as displayed in the UI. 8.5 MB, 12.3 MB
Size_bytes Float / Int (Derived) The file size converted to a standard numerical unit for analysis. 8500000, 12300000

Usage & considerations

Technical characteristics of
ultrasound (US)

Acoustic
principles

Ultrasound transducers emit high-frequency sound waves (2-18 MHz depending on application) that penetrate tissues and reflect back at interfaces between different tissue densities. The time delay and intensity of returning echoes are processed to create real-time images. Higher frequencies provide better resolution but limited penetration; lower frequencies penetrate deeper with reduced resolution.

Real-time
dynamic imaging

Ultrasound provides real-time video capture (15-60 frames per second), enabling assessment of organ motion, blood flow, fetal movement, and cardiac function. Cine loops can be recorded and reviewed frame-by-frame. This dynamic capability is unique among cross-sectional imaging modalities.

Doppler
imaging

Color Doppler and spectral Doppler modes assess blood flow direction and velocity using the Doppler effect. Color Doppler overlays flow information on grayscale anatomy (red/blue indicating direction). Pulsed-wave Doppler provides quantitative velocity measurements at specific locations. Power Doppler shows flow presence without directional information, more sensitive for small vessels.

Image quality factors

Operator-dependent technique significantly affects image quality. Transducer selection (frequency, footprint), patient positioning, acoustic windows, gain settings, and focal zone placement all impact diagnostic quality. Body habitus, bowel gas, and bone limit visualization in some patients. Harmonic imaging improves contrast resolution.

Safety and accessibility

No ionizing radiation makes ultrasound ideal for pediatric imaging, obstetrics, and repeated examinations. Portable systems enable bedside point-of-care ultrasound (POCUS). Lower cost than CT/MRI improves accessibility. Real-time guidance for procedures (biopsies, drain placement, vascular access).

Clinical applications

Obstetric imaging (fetal anatomy, growth assessment), abdominal organs (liver, gallbladder, kidneys, pancreas), vascular (carotid stenosis, DVT screening, aortic aneurysm), cardiac (ejection fraction, valve function), thyroid/breast nodules, musculoskeletal (rotator cuff, tendon tears), and procedural guidance.

Primary use cases

  • Training deep learning models for automated fetal biometry measurement and gestational age estimation in obstetric ultrasound.
  • Developing real-time organ segmentation and tracking algorithms for surgical guidance and radiation therapy planning.
  • Building AI systems for automated quality assessment of ultrasound acquisitions and standard view recognition.
  • Creating Doppler flow analysis tools for quantitative vascular stenosis assessment and hemodynamic calculations.
  • Developing video-based models that analyze cine loops for cardiac function quantification (ejection fraction, wall motion).
  • Training models robust to operator variability and image quality variations inherent to ultrasound imaging.

Unlock your true
speed to scale 

Accelerate what data and AI can do together.