Exploratory identification of image-based biomarkers for solid mass pulmonary tumors

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Current quantitative imaging metrics for cancer therapy response assessment are often size-based.
• These size-based metrics may not accurately reflect disease progression in advanced cancers.
• There is a need for reproducible and predictive radiologic biomarkers.

Purpose of the Study:

• To explore the utility of statistical textural features from medical images for cancer change detection.
• To develop an end-to-end process for analyzing image-based features beyond tumor size.
• To identify radiologic features that are more predictive of time-to-progression (TTP).

Main Methods:

• Utilized statistical textural features for cancer change detection.
• Employed the earth mover's distance metric to quantify changes in tumor burden between baseline and therapy response scans.
• Correlated imaging-derived change measurements with known patient time-to-progression (TTP) outcomes.

Main Results:

• Local texture energy emerged as a highly predictive feature for disease progression.
• This texture-based metric demonstrated superior predictive power compared to traditional tumor size measurements.
• The developed method offers a novel approach to assessing cancer therapy response.

Conclusions:

• Statistical textural features, particularly local texture energy, offer a more predictive biomarker for cancer progression than tumor size.
• This approach enhances the assessment of cancer therapy response and patient outcomes.
• The findings support the integration of advanced imaging analysis for improved cancer management.