Dermoid Cyst Detection on CT

Purpose	Detection and classification of markers used in mammography.
Tag(s)
Panel	Breast
Define-AI ID	23060009
Originator	Richard Foster
Lead	Richard Foster
Panel Chair	Bhavika Patel, Robert Rosenberg
Panel Reviewers	Breast Imaging Panel
License	Creative Commons 4.0
Status	Public Comment
RadElement Set	RDES269

Clinical Implementation

Value Proposition

Utilizing skin markers to communicate the location of potential benign findings in mammography has been a long-standing practice. A nipple out of profile, a post-surgical scar, and a raised mole can all cause interpretation concerns if not identified as such for the radiologist. The ability for the AI engine to identify the presence of a marker, the location of the marker, and the shape of the marker could be useful in the following ways:

Nipple marker.
- A nipple marker is useful to determine the location of the nipple when it is not evident on the exam.
- a nipple marker is helpful to differentiate a retroareolar lesion from a nipple out of profile.
- A nipple marker helps establish the nipple pectoral line when the nipple is not evident.
Mole marker.
- The mole marker is useful when the mole images as an interparenchymal density and it cannot be determined if that density is on the skin or within the breast.
Scar marker.
- The scar marker is useful to create correlation between a post-surgical scar and normal signs of scarring such as architectural distortion, calcifications, skin thickening, puckered or retracted skin and a spiculated mass.

Narrative(s)

A 75 year old female presents for screening mammography. The technologist observes a nipple out of profile, a raised skin mole or a post-surgical scar on the patient’s breast. According to the facility’s protocol, the technologist places a shaped marker as designated by the facility to communicate that finding. The nipple is marked with a pellet, a raised mole is marked with a lucent circle and the post surgical scar is marked with a lucent linear marker cut to the length of the scar. The technologist indicated on the worksheet that a marker was used and charts the location. The AI engine correctly identifies the shape of the marker and the location.

Workflow Description

The screening mammogram with the marker in place is reviewed by the AI algorithm. The AI engine correctly identifies the marker based on the:

The shape of the marker (Sphere, circle, linear)
Input from the technologist stating that a marker was placed on the breast

The AI engine can then communicate the location of the marker and the shape of the marker.

Considerations for Dataset Development

Procedure(s)	Screening Mammography
View(s)	CC and MLO screening views
Age	40+

Technical Specifications

Inputs

DICOM Study

Procedure	Screening Mammogram
Views	CC and MLO view
Data Type	DICOM
Modality	MAM
Body Region	Chest
Anatomic Focus	Breast

Primary Outputs

Skin Marker Detection

RadElement ID	RDE1718
Definition	Identify if skin markers are present.
Data Type	Categorical
Value Set	Unknown Marker present Marker absent
Units	N/A

Location of skin marker

RadElement ID	RDE1719
Definition	Identify the location of skin marker
Data Type	Coordinates
Value Set	NA
Units	N/A

Shape of skin marker

RadElement ID	RDE1720
Definition	Identify the shape of skin marker
Data Type	Categorical
Value Set	Nipple marker - pellet marker Mole marker - lucent circle marker Scar marker - lucent linear marker
Units	N/A

Laterality of skin marker use

RadElement ID	RDE1721
Definition	Was the marker used on the left or right breast
Data Type	Categorical
Value Set	Left Right
Units	N/A

Skin Marker on previous studies

RadElement ID	RDE1722
Definition	Identify if there are any skin markers on previous studies in the same location if applicable
Data Type	Categorical
Value Set	Yes No Unknown
Units	N/A

Future Development Ideas

Future development ideas include the potential of the AI engine to use the information that a specific marker was present in a specific location to enhance the final recommendation for that study.

This seems like a useful first step for markers and AI. It could help populate a report with the marker information.