Radiomic Analysis of CT Scans Can Predict Aggressiveness of Early Stage Lung Cancer

August 21, 2020

Specific measures of a tumor’s texture and intensity as they appear on CT images can help determine how much treatment patients with early stage lung cancer need, a study found.

Screening for early lung cancer may become more effective at predicting the aggressiveness of disease if new knowledge about how to interpret the images is applied, a team of researchers has found.

Scientists from Moffitt Cancer Center in Tampa, Florida, used radiomics — an algorithm-based analysis of the features of radiographic images — to find two major characteristics of lung tumors that can be employed to divide cases into three risk categories.

Advantages of the method include its noninvasive nature and the fact that it is conducted during screening that would occur anyway. In addition, it gives a clear picture of the aggressiveness of a whole tumor, rather than of only one section, as is the case when tissue is tested, according to the authors.

The study was published in Nature Scientific Reports.

Screening for lung cancer with low-dose computed tomography (CT) is available to people with certain histories of heavy smoking. The annual imaging is associated with a 20% reduction in lung cancer death, but it sometimes finds slow-growing tumors that are treated even if they never would have caused any harm to the patient; this can leave patients with unnecessary long-term side effects. The use of radiomics, according to a Moffit press release, may help identify the patients with early stage lung cancer who face a higher risk of poorer outcomes, allowing them to undergo aggressive follow-up or added therapy after primary treatment while those deemed at lower risk can skip those steps.

“Identifying predictive biomarkers that detect aggressive cancers or those that may be slow-developing and non-emergent are a critical unmet need in the lung cancer screening setting,” Matthew Schabath, one of the study’s authors and an associate member of Moffit’s cancer epidemiology department, said in the press release. “Additional research is needed to inform us on the potential translational implications of this model, but it could make a major impact on saving lives by identifying lung cancer patients with aggressive disease while also sparing others from unnecessary therapy.”

Study Design

In the study, the researchers used data from the previously conducted National Lung Screening Trial (NLST), which led to the widespread use of low-dose CT screening for lung cancer. Looking at data from patients in that trial whose screenings were positive for lung cancer, the scientists found radiomic features associated with the size, shape, volume and texture of tumors (intratumoral) and their environment (peritumoral) that were linked with a higher risk of aggressive disease. In addition, they compared the radiomic characteristics of high-risk tumors against previously conducted genomic analysis to determine which cancer-driving genetic mutations were associated with aggressive disease.

“Our goal was to use radiomic features to develop a reproducible model that can predict survival outcomes among patients who are diagnosed during a lung cancer screening,” Jaileene Pérez-Morales, lead study author and a postdoctoral fellow at Moffitt, said in the release.

The research team conducted its investigation in 161 patients from the NLST and validated the results in a separate group of 73 patients from that study; it also tested the results in 62 patients with lung cancer that was not detected through screening. All three groups were matched according to age, sex, smoking status, number of pack-years smoked, family history of lung cancer, treatment, stage, screening and the cancer’s histology— meaning its microscopic structure.

After analyses to remove redundant and non-reproducible radiomics features, the researchers came up with a predictive model relying on two radiomic features, one peritumoral and one intratumoral, to stratify patients into three risk groups: low, intermediate and high. The model was most useful in patients with early stage disease.

Findings

Across all patients, those in the high-risk group had an overall survival — meaning the length of time from diagnosis until death — five times worse than that in patients with low- and intermediate-risk disease; they also had a time until disease progression that was four times worse.

Most patients (92%) in the high-risk group were male, versus 54% in the low-risk group. Early stage disease was more common in those with low risk versus high-risk lung cancer (80% versus 33%).

The two radiomics features used to stratify the patients were neighborhood gray-tone difference matrix (NGTDM) busyness, a peritumoral textural feature, and statistical root mean square (RMS), an intratumoral intensity feature. RMS was significantly associated with expression of the protein FOXF2, with lower expression linked with shorter survival.There was also an association between RMS and LOC285043, an uncharacterized gene.

Finally, RABGAP1L was associated with NGTDM busyness, with the intermediate- and high-risk groups having a higher expression of the protein.


x