Assessing Response to Treatment in Patients With Solid Tumors

OBJECTIVES:

Primary

- To compare the volumetric response achieved using semi-automated computer techniques
with the unidimensional (RECIST) and bidimensional (WHO) criteria as part of clinical
trials running at MSKCC, testing new agents for different types of solid tumors.

Secondary

- To assess the effect of CT slice thickness on 3D, 2D, and 1D size measurements.

- To study variability (reproducibility) of the three segmentation algorithms developed
for liver, lung, and metastatic lymphadenopathy.

- To compare computer algorithm variability against intra- and inter-radiologist
variability.

- To validate each algorithm's accuracy by comparing measurements obtained from the
computer algorithm with the "Gold Standard" radiologist reading.

OUTLINE: The research project coordinator (RPC) is notified when a patient is scheduled for
CT scan. The RPC forwards this information to the CT scan technician performing the CT scan
on the patient per the therapeutic protocol. CT images of the patient reconstructed at
different slice thicknesses are electronically transferred from the hospital PACS to the
research PACS server where patient identification information is de-identified. The images
stored in the research PACS are accessed by the research workstations where image processing
and analysis is performed.

Segmentation algorithms are used to segment tumors, measure tumor sizes, and size changes in
the course of chemotherapy. The variability of the algorithm and effects of CT image slice
thickness on the size measurements is also studied.