肺扁平上皮癌におけるコンピュータ支援定量的間質解析：予後的意義に関する包括的調査

The tumor-stroma ratio (TSR) has been validated as an independent prognostic factor across multiple cancer types, including lung squamous cell carcinoma (LUSC), in which a stroma-rich phenotype is generally linked to poor outcomes. However, traditional manual TSR assessment is subjective and has limited reproducibility. This study aimed to investigate the clinical utility of computer-aided (CA) quantitative stromal analysis in LUSC, elucidate the correlation between TSR and clinicopathological features, and provide a theoretical basis for precision diagnosis and treatment.

There were 189 patients with primary LUSC diagnosed between 2015 and 2020 included in a retrospective cohort. Using automated analysis techniques in conjunction with whole-slide imaging (WSI), a quantitative evaluation of the TSR was carried out automatically. Meanwhile, the TSR was also evaluated manually. Based on the median TSR value, patients were categorized into cohorts that were either stroma-poor (TSR-L) group or stroma-rich (TSR-H) group. Using Cox regression models and Kaplan-Meier survival analysis, prognostic significance was tested. In addition, the consistency between the results of manual evaluation and those of CA evaluation was analyzed.

The median CA-TSR score was 0.34. Patients in the TSR-H group demonstrated significantly worse overall survival (OS) and disease-free survival (DFS) (both P<0.001), which corresponded with pleural invasion (P=0.03) and advanced pathological staging. The area under the curve (AUC) for OS demonstrated the superior predictive accuracy of CA-TSR over manual scoring (0.651-0.7 vs. 0.488-0.573), and multivariate analysis validated it as an independent prognostic factor [DFS: hazard ratio (HR) =2.790; OS: HR =2.633]. Furthermore, the manual and CA evaluations showed a moderate level of agreement (r=0.45, P<0.001).

CA-driven quantitative TSR analysis enables objective and efficient assessment of the stromal ratio in LUSC. The CA-TSR score serves as an independent prognostic predictor, significantly enhancing survival prediction accuracy compared to conventional methods.