Objective To discern highly aggressive intratumoral areas among lung adenocarcinoma (LUAD) and its impact on occult nodal metastases and the recurrence rate with radiomic analysis. Methods This prospective dual-institution study analyzed clinical information and high-resolution preoperative CT of 528 patients from institution A and 249 patients from institution B. We extracted radiomic features and performed pathologic evaluations for resected tumors, based on the 2020 International Association for the Study of Lung Cancer (IASLC) classification. Prediction models were developed to discern micropapillary and solid patterns within LUAD using clinical and radiomic features from institution A through logistic analysis. Results Six selected CT radiomic features, sex, CTR (consolidation-to-tumor ratio), and solid diameter were selected to develop the prediction models. A composite model of radiomic and clinical characteristics outperformed radiomics-only and clinical-only models (AUC, 95% CI; the composite model: 0.84 [0.81–0.87]; the radiomics model: 0.82 [0.78–0.87]; the clinical model: 0.80 [0.76–0.83]) in institution A. External validation was performed with institution B cohort, showing even better results (AUC, 95% CI; the composite model: 0.91 [0.87–0.94]; the radiomics model: 0.89 [0.84–0.94]; the clinical model: 0.88 [0.84–0.92]). Conclusions Our study underscores the potential of radiomics to preoperatively predict aggressive histologic patterns in LUAD, enabling precise treatment planning and prognosis estimation. Key Points <Emphasis Type="BoldItalic">Question</Emphasis> Can any adjuvant methods address the limitations of core needle biopsies, which are invasive and may not capture the full heterogeneity of lung adenocarcinoma? <Emphasis Type="BoldItalic">Findings</Emphasis> In a prospective study of 528 patients with cT1N0M0 lung adenocarcinoma, a composite model of clinical characteristics, conventional CT findings, and radiomics features predicted high-grade cancers. <Emphasis Type="BoldItalic">Clinical relevance</Emphasis> Preoperative non-invasive diagnosis of histologically high-grade tumors using radiomics analysis offers crucial information for the treatment of lung adenocarcinoma with respect to occult lymph node metastasis and recurrence rate.

Deciphering the intratumoral histologic heterogeneity of lung adenocarcinoma using radiomics

期刊论文
发表日期：

2025年2月12日

查看全文

Jae Mo Koo,
Jonghoon Kim,
Junghee Lee,
Soohyun Hwang,
Hyo Sup Shim,
Tae Hee Hong,
Yu Jin Oh,
Hong Kwan Kim,
Chang Young Lee,
Byung Jo Park,
Ho Yun Lee

摘要

Objective

To discern highly aggressive intratumoral areas among lung adenocarcinoma (LUAD) and its impact on occult nodal metastases and the recurrence rate with radiomic analysis.

Methods

This prospective dual-institution study analyzed clinical information and high-resolution preoperative CT of 528 patients from institution A and 249 patients from institution B. We extracted radiomic features and performed pathologic evaluations for resected tumors, based on the 2020 International Association for the Study of Lung Cancer (IASLC) classification. Prediction models were developed to discern micropapillary and solid patterns within LUAD using clinical and radiomic features from institution A through logistic analysis.

Results

Six selected CT radiomic features, sex, CTR (consolidation-to-tumor ratio), and solid diameter were selected to develop the prediction models. A composite model of radiomic and clinical characteristics outperformed radiomics-only and clinical-only models (AUC, 95% CI; the composite model: 0.84 [0.81–0.87]; the radiomics model: 0.82 [0.78–0.87]; the clinical model: 0.80 [0.76–0.83]) in institution A. External validation was performed with institution B cohort, showing even better results (AUC, 95% CI; the composite model: 0.91 [0.87–0.94]; the radiomics model: 0.89 [0.84–0.94]; the clinical model: 0.88 [0.84–0.92]).

Conclusions

Our study underscores the potential of radiomics to preoperatively predict aggressive histologic patterns in LUAD, enabling precise treatment planning and prognosis estimation.

Key Points

Question Can any adjuvant methods address the limitations of core needle biopsies, which are invasive and may not capture the full heterogeneity of lung adenocarcinoma?

Findings In a prospective study of 528 patients with cT1N0M0 lung adenocarcinoma, a composite model of clinical characteristics, conventional CT findings, and radiomics features predicted high-grade cancers.

Clinical relevance Preoperative non-invasive diagnosis of histologically high-grade tumors using radiomics analysis offers crucial information for the treatment of lung adenocarcinoma with respect to occult lymph node metastasis and recurrence rate.