Evaluating Size of Lung Masses

Purpose	The increase, decrease, or stability of a known primary tumor is the starting point of all cancer staging exams. The goal would be to develop an algorithm that given a cancer study could determine the dominant lesion. Once that lesion is identified on the current study, It should be found on a comparison study and the volumes should be compared.
Tag(s)
Panel	Thoracic Panel
Define-AI ID	19080007
Originator	Christopher P. Gange
Lead	Christopher P. Gange
Panel Chair	Eric Stern
Panel Reviewers	Thoracic Panel
License	Creative Commons 4.0
Status	Published
RadElement Set	RDES82

                

Clinical Implementation

Value Proposition

Staging cancer accurately is a common and time consuming task in radiology, and lung cancer is the most common lethal malignancy in the world. An algorithm that automated these comparisons would improve radiologist’s workflow, and would hopefully decrease variability in reporting. This would allow oncologists to make better treatment decisions for cancer patients.

Narrative(s)

A patient being treated with a targeted therapy for lung cancer goes for their routine staging scan before a clinic visit. The algorithm flags this study as a significant increase in tumor volume, alerting the radiologist to read the study more promptly and look for signs of metastatic disease. The accurate staging allows the patient to get optimized oncology care.
A patient undergoing chemotherapy goes for their routine staging scan. The algorithm allows a quick read and the patient receives the good news of a treatment response before they leave the clinic.

Workflow Description

The image is obtained from modality and sent to PACS and the AI engine. The image is analyzed by the engine. The system then detects the primary tumor and reports its size. Size measurements are generated on all subsequent CT studies to determine how the tumor size is changing over time. An alert message is sent to PACS from the engine with the size information, identification, and graphic highlighting the tumor and its change.

Considerations for Dataset Development

Age	Varied
Sex at birth	Male, Female
Tumor size	Varied
Smoking history	Yes, No
Chronic obstructive pulmonary disease	absent, present
Diffuse lung fibrosis	absent, present
Tumor histology	adenocarcinoma, squamous cell, Large cell, small cell, carcinoid

Technical Specifications

Inputs

 

Current Chest CT

Primary Outputs

Change in mass size

Current tumor volume

Tumor volume change

Tumor max diameter