Digital Pathological Image Analysis Predicts Aggressive Lung Cancers

Quantitative morphological features of tumor pathological images may help predict prognosis in patients with non-small cell lung cancer.

Quantitative morphological features of tumor pathological images may help predict prognosis in patients with non-small cell lung cancer (NSCLC), according to Texas researchers. Currently, only a small percentage of the total morphological features are analyzed when determining treatment. However, it may be possible to change that.

After evaluating more than 900 differences in the shape and structure of cancer cells, researchers developed objective and quantitative computational approaches to analyze the morphological features of pathological images for patients with NSCLC. They report in the March edition of the Journal of Thoracic Oncology that using a new algorithm it may be possible to better stratify NSCLC patients into low-risk and high-risk categories.  

“This computational approach should someday make it possible for doctors to tailor the treatment of individual patients based on risk predicted by computer algorithms, for instance choosing to treat patients at higher risk more aggressively,” said senior study author Guanghua Xiao, PhD, who is an Associate Professor of Clinical Sciences and Bioinformatics at UT Southwestern Medical Center in Dallas.

The researchers analyzed 3,206 slides containing pathological images of cancer tissue from 523 patients with adenocarcinoma and 511 patients with squamous cell carcinoma. The slides came from The Cancer Genome Atlas, which was started more than a decade ago to catalog genetic, image, and clinical aspects of cancers.

The researchers sorted the slides for features involving cell size, shape, distribution, texture, and location of nuclei. They were able to identify 12 to 18 features that could divide the cancers into aggressive (high-risk) and less aggressive (low-risk) groups. For both adenocarcinoma and squamous cell carcinoma, patients in the high-risk group had more than twice the risk of death as those identified as low-risk when using this computational approach, according to Xiao.

This predictive model could help improve current clinical practice and assist pathologists and clinicians in diagnosis and decision-making for patients with lung cancer. However, Xiao said before this computerized image analysis could be used clinically, it would need to be tested in a prospective study evaluating the effects of different image resolutions, specimen sizes, and types of samples.

In recent years, studies have suggested that the heterogeneity of tumor cells and close interactions between tumor cells and their microenvironments may play a much more important role than previously recognized. Now, investigators hoped to capitalize on this new knowledge. While this study only looked at patients with adenocarcinoma and squamous cell carcinoma, this quantitative computational approach may have applicability in a host of tumor types.