Pediatric Hip US - relatively simple classification scheme

 

Purpose

Automated evaluation of hip ultrasound studies to aid in screening, initial evaluation, and follow-up of developmental dysplasia of the hip (DDH).

Tag(s)

 

Panel

Pediatric

Define-AI ID

20010008

Originator

Safwan Halabi
Lead Rahul Nikam

Panel Chair

Steven Blumer

Panel Reviewers

Morgan McBee, Viet Le, Neil Lall, Safwan Halabi, and Steven Blumer

License

Creative Commons 4.0
Status Public Comment
RadElement Set RDES182

 
   
   
   
   
   
                               

Clinical Implementation


Value Proposition

Developmental dysplasia of the hip (DDH) encompasses a broad spectrum of abnormal hip development during infancy and early development. The definition encompasses a wide range of severity, from mild acetabular dysplasia without hip dislocation to frank hip dislocation, and is the most common pediatric hip disorder, affecting 1 - 3% of all infants. When left untreated or missed during early screening, DDH can lead to debilitating long term complications, including early-onset osteoarthritis of the hip, pain, limping, and the need for a total hip replacement in early adulthood.

Imaging plays a vital role in screening, initial diagnosis and further workup of DDH. Different screening strategies have been described, including clinical examination alone, selective ultrasound screening and universal ultrasound screening. Accurate diagnosis and classification of DDH requires expertise of a specialized pediatric orthopedic surgeon and pediatric radiologist. Nationwide, a significant shortage of well-qualified pediatric radiologists exists, as there are approximately 950 pediatric radiologists nationally which comprise 3.5% of practicing US radiologists, limiting access to care.. To be adopted in clinical practice, Artificial Intelligence (AI) applications must address unmet needs or improve on existing solutions. Given the dearth of specialists, AI techniques show promise to be incorporated in the clinical workflow of studies performed for DDH diagnosis and screening, or as an add-on to the radiologist’s interpretation for DDH studies.


Narrative(s)

A 10-week-old female infant with a history of breech presentation at birth, and equivocal findings of hip instability on physical examination of the hip. Ultrasound of the hip is performed to evaluate developmental dysplasia of the hips. An algorithm evaluates the images, classifies the hip, and sends a message to the PACS. 


Workflow Description

The images are obtained from the ultrasound modality and sent to PACS and the AI algorithm or model. The algorithm or model analyzes the images, and detects and classifies the abnormality, and sends a message to the PACS. 


Considerations for Dataset Development



Procedure(s)

US, Hips

Sex at birth

Male, Female

Age

4 weeks to 6 months

Indication

Evaluation of symptomatic child, screening, follow up

Risk factors

Breech presentation, family history of parent and/or sibling with DDH, neuromuscular pathologies, treatment monitoring, oligohydramnios, asymmetric skin folds in the medial thigh

Technical Specifications


Inputs


DICOM Study

Procedure

US, both hips

Views

6 views, coronal neutral, coronal flexion, transverse flexion in adduction, transverse flexion in abduction, coronal flexion posterior acetabulum without and with stress

Data Type

DICOM

Modality

US

Body Region

Hip

Anatomic Focus

Hip Joint





Primary Outputs


Detection of femoral head position

RadElement ID

RDE1258

Definition

Detection of femoral head position

Data Type

Categorical

Value Set

  • Normal

  • Subluxation

  • Dislocation


Units

N/A



Stability

RadElement ID

RDE1259

Definition

Stability

Data Type

Categorical

Value Set

  • Stable

  • Unstable

Units

N/A




Acetabular Morphology

RadElement ID

RDE1260

Definition

Acetabular Morphology

Data Type

Categorical

Value Set

  • Normal

  • Immature

  • Dysplastic


Units

N/A


Alpha angle

RadElement ID

RDE1261

Definition

Alpha angle

Data Type

Numerical

Value Set

N/A

Units

Degrees





Secondary Outputs


Beta angle

RadElement ID

RDE1262

Definition

Beta angle

Data Type

Numerical

Value Set

N/A

Units

Degrees


Femoral head coverage

RadElement ID

RDE1263

Definition

Detection of femoral head coverage

Data Type

Categorical

Value Set

  • > 50 % Coverage

  • < 50 % Coverage



Units

N/A


Future Development Ideas


Algorithms that utilize 3D ultrasound data obtained in rest and stress, instead of 2D images in coronal and transverse planes and algorithms that perform automated acetabular segmentation in addition to classifying hip abnormalities. This approach will significantly eliminate the need for dependence on operator expertise for image acquisition. 

Deployment of autonomous AI-directed ultrasound system for imaging and classification of DDH.