Diverticulitis Detection

Purpose	To detect colonic diverticulitis on CT
Tag(s)
Panel	Abdominal
Define-AI ID	19020005
Originator	Luther B Adair, II, MD
Lead	Luther B Adair, II, MD
Panel Chair	Arun Krishnaraj, MD
Panel Reviewers	Abdominal Panel
License	Creative Commons 4.0
Status	Public Comment
RadElement Set	RDES138

Clinical Implementation

Value Proposition

Colonic diverticulitis is an acute infection and common cause of emergency visits. AI can creatve value to a variety of stakeholders by detecting, quantitating, comparing, and alerting both nonspecialists and radiologists to this abnormality. In clinical settings where a radiologist is not readily available such detection would be benefit non-radiologists such as technologists who perform the CT and alert them if diverticulitis is present. 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. One possible extension to this use case is to detect diverticulitis on CT and assess the need for further treatment based on the absence or presence of complications.

Narrative(s)

A 58-year old patient presents to an urgent care facility with worsening left lower quadrant pain, loose stools and fever and is taken to the emergency room at which time a CT scan is ordered.

Workflow Description

The relevant images are obtained from the 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 or measurements. An alert messages sent it back to the engine with information, identification, and graphics highlighting the wall thickening, pericolonic infiltrative changes, pericolonic lymph nodes, extramural fluid, or free air.

An algorithm evaluates the CT scan and categorizes whether colonic wall thickening is present, absent, or undetermined. If present, it then assesses whether pericolonic infiltrative changes are present, absent, undetermined; assesses whether extramural fluid present, absent, or undetermined; assesses whether pericolonic lymph nodes are present, absent, or undetermined; whether an organized extramural fluid is present, absent, or undetermined; assesses whether free air is present, absent, or undetermined. For cases returned where the algorithm returns colonic wall thickening is present and certain secondary output elements are present (lymph nodes present and pericolonic infiltrative changes not present), the radiologist is made aware to look for malignancy and/or diverticulitis. For cases returned where the algorithm returns colonic wall thickening is present and more than 2 secondary inputs are present (I.e pericolonic infiltrative changes present and extramural fluid present), the exam is prioritized in the radiologist’s worklist for urgent interpretation and reporting.

Considerations for Dataset Development

Procedure(s)	{CT Abdomen, with contrast, without contrast, CT Abdomen and pelvis}
View(s)	{sagittal, axial, coronal} Clinical Note: Subtle colonic wall thickening can be overlooked
Sex at Birth	{Male, Female}
Age	[20,90]
Abdominal pain	{left lower quadrant pain}
Percutaneous Tube	{ percutaneous drainage tube}
Other	{extramural abscess, colonic malignancy, colonic perforation}

Technical Specifications

Inputs

 

DICOM Study

Primary Outputs

Sigmoid Colonic wall thickness

Secondary Outputs

Detection of pericolonic infiltrative changes

Pericolonic lymph node detection

Extramural Fluid Detection

Extramural Fluid Classification

Detect Free Air

Related Datasets

The Cancer Imaging Archive (TCIA)

• C4KC-KiTS 
• CPTAC-CCRCC 
• QIN Prostate 
• Prostate-3T 
• Prostate-Diagnosis 
• Prostate-MRI 

Know of more related datasets? Please let us know