Compression Fracture Detection

Purpose

Detect and describe spinal compression fractures

Tag(s)

 

Panel

Neuroradiology Panel

Define-AI ID

19020012

Originator

Ayis Pyrros, Adam Flanders
Lead Ayis Pyrros, Adam Flanders

Panel Chair

Alex Norbash

Panel Reviewers

Neuroradiology Panel

License

Creative Commons 4.0 
Status Public Commenting 
RadElement Set RDES90 
                               

Clinical Implementation


Value Proposition

Spinal compression fractures can be a potentially life-threatening and readily treated emergency. AI meeting this use case would help in detecting, quantitating, comparing, and alerting both non-specialists and radiologists to this potentially life-threatening abnormality. In clinical settings where a radiologist is not readily available, such detection would be of value to non-radiologists. In clinical settings where radiologists are available, such detection could be used to alert the radiologist and prioritize its reporting and notification to the treating physicians. Possible extensions to this use case is to track changes in compression fractures, identify other types of compression fractures, and to attempt to categorize fractures using AO criteria regarding stability.

Narrative(s)

25 year old presents after a motor vehicle accident and lumbar radiographs are taken in the emergency room. The algorithm evaluates image and categorizes the spine as compression fracture present, absent, or undetermined. For cases when the algorithm returns compression fracture present, if a radiologist is not present at the time of imaging, an alert is provided to the ordering physician. If a radiologist is present, the exam is prioritized in the radiologist worklist for urgent interpretation and reporting.
A 70 year old presents after a fall to the emergency room with back pain. Algorithm evaluates image and categorizes the spinal radiographs as being normal, abnormal or undetermined, for compression fractures. For cases when the algorithm returns compression fracture present, if a radiologist is not present at the time of imaging, an alert is provided to the ordering physician. If a radiologist is present, the exam is prioritized in the radiologist worklist for urgent interpretation and reporting (particularly for unstable fractures, using the AO criteria).

Workflow Description

Image obtained from modality and sent to PACS and the AI engine. Image analyzed by engine. System detects fractures and stability. An alert message is sent to PACS from the engine with the information, and identification, and graphic highlighting the possible spinal compression fractures.

Considerations for Dataset Development



Procedures(s)


{XRAY, Lumbar, Thoracic, and cervical; CT, , thoracic, and lumbar}

View(s)

{AP, PA/Lat, flexion/extension}

Clinical Note: Many compression fractures are chronic or subtle. In addition, osteopenia can make detection difficult.

Sex at Birth

{Male, Female}

Age

[0,100]

Spinal Trauma (or intervention)

{fall, MVC, direct trauma}

Comorbidities

{unstable, retropulsion, spinal hematoma, spinal cord injury)

Spinal Involvement

{Cord injury, spinal hematoma}

Other

{Osteopenia, pathologic fractures secondary to a bone lesion}

Clinical Note: Distinguishing acute and chronic on radiographs is very difficult in the absence of comparison studies.

Technical Specifications


Inputs

 

DICOM Study

 

Procedure

X-ray, lumber, thoracic, and cervical

Views

AP, PA/Lat, Flexion/extension

Data Type

DICOM

Modality

X-ray

Body Region

Cervical, thoracic, lumbar

Anatomic Focus

Spine



Primary Outputs


Fracture Detection

RadElement ID

RDE531

Definition

Fracture detection

Data Type

Categorical

Value Set

0- Unknown

1- Fracture

2- No Fracture

Units

N/A


Level of greatest compression

RadElement ID

RDE532

Definition

Level of greatest compression

Data Type

Categorical

Value Set

= C1

= C2

= C3

= C4

= C5

= C6

= C7

= T1

= T2

= T3

= T4

= T5

= T6

= T7

= T8

= T9

= T10

= T11

= T12

= L1

= L2

= L3

= L4

= L5

Units

N/A

Future Development Ideas


  • Algorithm compares prior imaging and returns differences in output elements (interval new fracture, or progressive fracture).