Background:

Lung adenocarcinoma is the most common lung cancer and has variable biology that affects outcomes and treatment. Numerous prognostic factors can be obtained from resected tumors, but a tool that could predict lung adenocarcinoma biology pre-operatively would help tailor decision-making.

SILA is a radiomic tool that assigns scores to 9 groups of CT features which are collectively used to calculate a risk that a tumor recurs. While SILA recurrence risks have been validated, it is not yet known how the 9 groups of features themselves might relate to lung adenocarcinoma characteristics.

In this study, our aim was to elucidate novel associations between the 9 scores and known prognostic factors in lung adenocarcinoma.

Methods:

We retrospectively reviewed resected lung adenocarcinoma cases with available preoperative CT images at our hospital between August 2016 and August 2022.

SILA analysis involved manual segmentation of lung lesions on CT scans. The software then generated scores for each of the 9 groups of features. The 9 features are referred to by color names, such as violet (V), indigo (I), etc. We reviewed the charts to record pertinent clinical-pathologic variables. Statistical tests used included chi square, Kruskal-Wallis to Dunn-Bonferroni-Tests post-hoc analysis, and Mann-Whitney U-Test.

Results: Study cohort included 276 patients with a median age of 69. There were 176 (63.4%) female patients and 197 (71.4%) stage I. There were 54 (19.6%) recurrences.

Our results showed positive and negative associations between certain color scores and markers of tumor aggression, including standardized reuptake value (SUV) on T1 scans (p-values ranging from <0.001 to 0.009) and tumor size on CT (p-values from <0.001 to 0.009). In addition, color scores had associations with known prognostic factors that can only be determined after resection, including visceral pleural invasion (p-values from <0.001 to .045) and lymphovascular invasion (p values from <0.001 to .037).

Color scores were also significantly different between certain pathologic stages. For example, violet values were significantly different between stage IB and IA2 tumors (adjusted p=0.001) and between stage IA2 and IIB tumors (adjusted p<0.001).

Conclusion:

The associations between SILA color scores and validated markers of tumor aggression point to the exciting possibility that color scores may provide information about lung adenocarcinoma biology pre-operatively. Future work will explore additional correlations between color scores and other pre- and post-operative clinicopathologic data.