Shunt Detector


Detect if the ventricular shunt catheter is intact and does not contain kinks




Pediatrics Panel

Define-AI ID



Morgan McBee
Lead Morgan McBee

Panel Chair

Matt Hawkins

Panel Reviewers

Pediatrics Panel


Creative Commons 4.0 
Status Public Commenting 
RadElement Set RDES118 

Clinical Implementation

Value Proposition

Hydrocephalus is a buildup of cerebrospinal (CSF) fluid within the ventricles of the brain. A common treatment is placement of a ventricular shunt catheter into the ventricles with catheter tubing extending under the patient’s skin and terminating most commonly in the peritoneal space within the abdomen (ventriculoperitoneal), but it can also extend into the right atrium of the heart (ventriculoatrial) or pleural space in the chest (ventriculopleural). Shunt catheter discontinuity or kinking can lead to shunt malfunction, return of hydrocephalus, and neurologic changes in children. Shunt catheter discontinuity is fairly uncommon and can often be quite subtle. As such, it can easily be overlooked. AI meeting this use case would help to reduce false negative rate, patient risk, and medical legal risk for the radiologists. Additionally, programmable valves are used to regulate the pressure of the shunt catheter system. There are many different brands and types of valves, each of which has its own guides for settings.


A 7-year-old patient with a history of shunted hydrocephalus presents to the emergency department with headaches, seizures, and vomiting. An emergent CT is performed which shows increased size of the ventricles (ventriculomegaly) which is concerning for shunt malfunction. A shunt series is obtained. The algorithm evaluates the continuity of the radiopaque portion of ventricular shunt catheter and identifies any regions where it is kinked and/or discontinuous, compares to prior studies for presence of old, abandoned elements of prior ventricular shunt tubing, evaluates the setting of the programmable shunt, and determines the tip location. The radiologist is informed of these data elements at the time of interpretation.

Workflow Description

Image obtained from modality and sent to PACS and automatically routed to the AI engine as a Shunt Series. The images are analyzed by the engine, and the system categorizes the type of shunt(s) (e.g., ventriculoperitoneal or ventriculopleural), number of valves, valve type (e.g., Codman), number of catheters, and then classifies the shunt catheter as either intact, discontinuous, or kinked, determines the programmable shunt valve setting(s), and identifies the location of tip of the catheter(s). If the catheter is not intact, the location information can also be sent to PACS to highlight the region the engine identified. A message is sent to PACS from the engine with the information.

Considerations for Dataset Development


All Ages


X-Ray, Shunt Series


Need to account for different views of the catheter. It may appear kinked on one view but is actually confirmed to be intact on the other.


The tip of the shunt catheter should be identified and should not be mistaken for a disruption.

Non-radiopaque portions

The algorithm should be able to infer whether the non-radiopaque portions of the shunt properly line up. This may actually be more easily overlooked by a radiologist than the radiopaque portion.

Abandoned shunt catheters

Sometimes portions of non-functioning shunt catheters are left inside the patient. The algorithm should be able to account for this and not mistake it for a disruption of a functioning catheter. Comparison to prior imaging may be helpful.

Programmable valve/type

Algorithm should be trained against a diverse set of programmable valve types and brands for clinical utility

Technical Specifications


DICOM Study 


X-ray, shunt series


AP and lateral of the skull, AP of the chest, abdomen

Data Type




Body Region

Head, Chest, Abdomen

Anatomic Focus



Primary Outputs 

Shunt Catheter Status

RadElement ID



The status of the VP Shunt

Data Type


Value Set

  • Unknown

  • Intact

  • Discontinuous

  • Kinked

  • Discontinuous and kinked



Shunt Tip Location

RadElement ID



The location of the catheter tip

Data Type


Value Set




Programmable Valve Setting

RadElement ID



The setting of the programmable valve

Data Type


Value Set



cm H20

Future Development Ideas

  • The algorithm could also run on other types of radiographic studies such chest, abdomen, pelvis, neck, skull, and spine studies, but the entire shunt would not be imaged.

Related Datasets

No known related public datasets at this time,  please us alert us if you know of any.