01
Ultrasound image
Newborn hip ultrasound image is transferred to the system for evaluation.
GrafMesh AI
Graf hip ultrasonography · Newborn hip screening
AI support for Graf ultrasound evaluation.
Capturing the standard plane correctly is the most critical and error-prone step in hip ultrasonography. GrafMesh AI identifies anatomical landmarks, automatically calculates α and β angles, and presents measurements alongside Graf classification.
GrafMesh AI·Graf ultrasound analysis
Product supported by the TÜBİTAK 2232-A International Leading Researchers Programme
End-to-end process
From ultrasound image to Graf type, in a single flow.
Graf evaluation begins with determining whether the image is in a standard plane suitable for measurement. Anatomical landmarks are then identified, α and β angles are calculated, and the resulting measurements are mapped to Graf classification. GrafMesh AI executes all these steps automatically.
02
Usability check for measurement
The straight appearance of the iliac wing, lower limb of the ilium, and labrum are evaluated to determine the image's suitability for Graf measurement.
03
Recognition of anatomical structures
The iliac wing, labrum, and necessary acetabular references are automatically identified by artificial intelligence.
04
Calculation of α and β angles
α and β angles are automatically calculated based on the lines established from anatomical landmarks.
05
Graf classification and results
The Graf type is determined by evaluating the measurements and relevant clinical parameters, and a summary of results is presented.
Standardization in Graf assessment
Reliable Graf evaluation relies on three fundamental steps.
Selecting a measurable standard plane, accurately identifying anatomical landmarks, and reducing inter-observer variability are essential for the consistency of Graf measurements. GrafMesh AI supports this process with shared anatomical and measurement references.
01
Measurable standard plane
The straight appearance of the iliac wing, lower limb of the ilium, and labrum are evaluated to determine the image's suitability for Graf measurement.
02
Accurate recognition of anatomical structures
The iliac wing, labrum, and acetabular references used in Graf measurement are automatically identified.
03
Inter-observer variability
Graf measurements can be influenced by operator experience. GrafMesh AI supports a more consistent approach across examiners by ensuring the use of standardized anatomical and measurement references.
Graf Classification
From measurement to Graf classification.
Alpha (α) and Beta (β) angles calculated from anatomical landmarks are evaluated alongside age information to determine the Graf type.
NormalMild dysplasiaSevere dysplasia
Type I
NormalNormal
α ≥ 60°
Mature hip morphology.
Type IIa / IIb
MildPhysiological immaturity · Mild dysplasia
α 50°–59°
The distinction between IIa and IIb is evaluated alongside the infant's age, maturation status, and other Graf criteria.
Type IIc / D
SevereDysplastic / instability risk group
α 43°–49°
Supports identifying cases requiring advanced specialist evaluation in conjunction with measurement results and additional Graf criteria.
Areas of use
From clinical evaluation to training, the same Graf approach.
GrafMesh AI delivers a reliable, standardized, and fast decision-support layer in every clinical setting where newborn hip screening is performed.
Routine screening
ScreeningSupports determining the standard plane suitable for measurement in newborn hip ultrasonography.
Clinical decision support
Decision SupportAI support for orthopedics, radiology, and relevant clinical specialties in standard plane evaluation, anatomical landmark identification, and angle measurement.
Specialist training
TrainingA supportive educational tool for resident candidates and healthcare personnel learning the Graf method, teaching the correct section and anatomical landmarks with real-time feedback.
Scientific foundation
From academic research to GrafMesh.
The scientific foundation of GrafMesh is based on published peer-reviewed studies on determining measurable standard planes, recognizing anatomical structures, α–β angle measurement, and Graf classification in newborn hip ultrasonography.
Real clinical images · Expert references · Peer-reviewed scientific studies
TÜBİTAK 2232-A support
GrafMesh development efforts are supported under the TÜBİTAK 2232-A International Leading Researchers Programme.
Independent clinical validation
Automated analysis results were compared against reference evaluations by experienced specialists on real newborn hip ultrasound images.
Full compliance with Graf standard
Measurable standard plane, anatomical references, and α–β angle geometry are evaluated based on the defined anatomical and measurement principles of the Graf method.
Scientific Publications
GrafMesh's scientific journey.
GrafMesh AI is built upon a continuous research foundation ranging from anatomical structure recognition to automated classification and measurable standard plane identification in Graf hip ultrasonography.
2019·Applied Soft Computing
Automatic segmentation and classification of neonatal hips according to Graf's sonographic method: A computer-aided diagnosis system
Hasan Basri Sezer · Aysun Sezer
Early foundational study combining automated anatomical structure identification, α–β angle measurement, and classification in a unified computer-aided workflow according to the Graf method.
164 hip ultrasound images
2020·Ultrasound in Medicine & Biology
Deep Convolutional Neural Network Based Automatic Classification of Neonatal Hip Ultrasound Images: A Novel Data Augmentation Approach with Speckle Noise Reduction
Aysun Sezer · Hasan Basri Sezer
Presents automated AI classification of newborn hip ultrasound images in the Graf standard plane and a novel data augmentation approach based on speckle noise reduction.
675 hip ultrasound images · 97.70% accuracy
2023·Joint Diseases and Related Surgery
Segmentation of measurable images from standard plane of Graf hip ultrasonograms based on Mask Region-Based Convolutional Neural Network
Aysun Sezer · Hasan Basri Sezer
A validated segmentation study focused on determining the measurable Graf standard plane through simultaneous automated evaluation of the iliac wing, labrum, and acetabular region.
675 hip ultrasound images · ~97% segmentation performance