Pancreatic cyst (< 1.5 cm) detection on CT

Purpose

To detect pancreatic cysts measuring less than 1.5 cm on CT and provide follow up recommendations based on age.

Tag(s)

 

Panel

Abdominal                                   

Define-AI ID

21020016

Originator

Luther B. Adair, II  
Lead Luther B. Adair, II and Kiran Thakrar

Panel Chair

Luther B. Adair, II  

Panel Reviewers

Andrew Smith

License

Creative Commons 4.0

Status Public Comment
RadElement Set
RDES193 

 
   
   
   
   
   
   
   
   

 
   
   
   
   
   
   
   
   
       

Clinical Implementation


Value Proposition


Incidentally discovered pancreatic cysts can be a challenging diagnostic conundrum for radiologists.  Standardizing the recommendations for followup and workup of these findings serves to help guide clinicians and ordering providers in a more reliable way and can potentially allow a data set to be created that allows research committees to modify the guidelines around incidentally discovered pancreatic cysts.  Upon discovering a pancreatic cyst, these guidelines proposed in the article Management of Incidental Pancreatic Cysts:  A White Paper of the ACR Incidental Findings Committee, 2017 will serve to offer a standardized and efficient approach to categorizing the incidental pancreatic cyst and the recommendations for follow-up .



Narrative(s)


A 60-year old patient presents with constant right upper quadrant pain and is taken to the emergency room at which time a CT scan of the abdomen with contrast is ordered.   An 8 mm cyst is incidentally found on the pancreas.


A 67-year old patient presents to an outpatient clinic with epigastric pain and is sent to the emergency room, at which time a CT scan of the abdomen and pelvis with contrast is ordered.  A 1.0 cm cyst is incidentally found on the pancreas. 


Workflow Description


The relevant images are obtained from modality and sent to PACS and the AI engine based on anatomic landmarks.  The images are analyzed by the engine.  The system then detects the relevant outputs providing associated categorical data, density, and measurements.  An alert message is sent to PACS from the engine with the information, identification, and graphics highlighting the pancreatic lesion, the density of the pancreatic lesion, and if it is a cyst measuring less than 1.5 cm, the recommendations for follow-up.  


An algorithm evaluates the CT scan and categorizes whether a pancreatic lesion is present, absent, or undetermined. If present, it then measures the density of the lesion in Hounsfield units; assesses whether it is a cyst with Hounsfield units measuring less than 20, a solid lesion with Hounsfield units measuring greater than 30, or indeterminate with Hounsfield units between 20 and 30.  If the lesion is solid or indeterminate, then additional imaging is recommended.  If a cyst is determined to be present, the algorithm measures the size.  If the size is less than 1.5 cm, the algorithm adjusts the recommendations based on age.  If the patient is less than 65 years old, a recommendation to reimage every year for 5 years is made to assess for stability.  If the patient is 65-79, a recommendation to reimage every 2 years for a total of 5 years is made to assess for stability.   These recommendations are first suggested to the radiologist for sign off and once the case is approved, they are available to the ordering provider within the report.  These recommendations are based on the article, Management of Incidental Pancreatic Cysts:  A White Paper of the ACR Incidental Findings Committee, 2017.  



  

Considerations for Dataset Development


Procedures

{CT Abdomen with contrast, CT Abdomen and pelvis with contrast}

View(s)

{Axial, Coronal, Sagittal}

Age

[0,79]

Sex at Birth

{Male, Female}

Other

{pancreatic ductal dilatation, pancreatic mass, pancreatic stent/drain, Whipple, artifact from adjacent surgical clips, pancreatitis with peripancreatic fluid, pseudocysts}





Technical Specifications


Inputs


DICOM Study

Procedure

CT Abdomen with contrast, CT Abdomen without contrast, CT Abdomen and Pelvis with/without contrast

Views

axial, coronal, sagittal

Data Type

DICOM

Modality

CT

Body Region

Abdomen

Anatomic Focus

Pancreas





Primary Outputs


Pancreatic Lesion Detection

RadElement ID

RDE1281

Definition

Detection of lesion

Data Type

Categorical

Value Set

  • Absent

  • Present

  • Undetermined

Units

N/A

 

Pancreatic Lesion Measurement

RadElement ID

RDE1282

Definition

If a pancreatic lesion is present then determine maximal size measurement.

Data Type

Numeric

Value Set

N/A

Units

mm


Pancreatic Lesion Density Measurement

RadElement ID

RDE1283

Definition

If a pancreatic lesion is present then determine density in Hounsfield units.

Data Type

Numeric

Value Set

N/A

Units

HU (Hounsfield units)


Pancreatic Lesion Categorization

RadElement ID

RDE1284

Definition

Categorization of cyst if lesion attenuation is less than or equal to 20 HU, categorization as undetermined if lesion attenuation is 21-29 HU, and categorized as solid if lesion attenuation is greater than or equal to 30 HU.

Data Type

Categorical

Value Set

  • Absent

  • Present

  • Undetermined

Units

N/A




Secondary Outputs


Pancreatic Cyst Measuring <1.5cm Present

RadElement ID

RDE1285

Definition

Detection of cyst measuring less than 1.5 cm

Data Type

Categorical

Value Set

  • Absent

  • Present

  • Undetermined

Units

N/A


Age based Recommendations 

RadElement ID

N/A

Definition

If the patient is less than 65 years old, a recommendation to reimage every year for 5 years is made to assess for stability. If the patient is 65-79, a recommendation to reimage every 2 years for a total of 5 years is made to assess for stability.

Data Type

N/A

Value Set

Age

Units

N/A

Future Development Ideas



  • Further evaluation of the solid pancreatic lesion with follow up recommendations. 

  • Evaluation of the peripancreatic fat for stranding to assess for pancreatitis and peripancreatic collections.

Related Datasets


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