Journal of neuro-oncology · 2026
Predictive modeling of lung cancer subtype with integration of clinical and thoracic imaging features to guide early brain metastasis management.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Park Jino, Buclez Paul, Lubisich Jessica, Hanubal Krishna, Inda Andrea, Hui Caressa, Harris Jeremy, Simon Aaron
Аннотация
PURPOSE: Optimal treatment for synchronous brain metastases at lung cancer diagnosis depends on subtype, but histologic/molecular profiling is often delayed. We developed machine-learning models using clinical and imaging data to classify lung cancer into three subtypes with distinct brain metastasis treatment implications: small cell lung cancer (SCLC), EGFR-mutated non-small cell lung cancer (EGFR+), and EGFR-wild-type non-small cell lung cancer (EGFR-).
METHODS: We retrospectively identified patients with metastatic lung cancer diagnosed from 2016 to 2025 at a single institution. Patients without brain metastases at diagnosis formed the training cohort, while patients presenting with synchronous brain metastases formed the testing cohort. Clinical variables and thoracic imaging features from diagnostic chest CT reports were used to train LASSO logistic regression and random forest models, evaluated via AUC and standard diagnostic metrics. Feature contributions were assessed using logistic regression coefficients and random forest permutation importance.
RESULTS: Of 305 patients, 182 were used for training and 123 comprised the testing cohort. Logistic regression achieved AUCs of 0.887, 0.811, and 0.801 for EGFR+, SCLC, and EGFR-, respectively; random forest achieved 0.907, 0.783, and 0.810. Adding imaging features to clinical variables improved prediction for EGFR + and SCLC. Contributory imaging features included fibrosis, emphysema, miliary pattern, cavitation, pleural attachment, and vessel encasement, among others. Three-way prediction showed highest positive predictive value for EGFR + and high negative predictive value for SCLC.
CONCLUSION: Combined clinical and chest imaging features can help predict lung cancer subtype, supporting early multidisciplinary discussions for patients with synchronous brain metastases.
METHODS: We retrospectively identified patients with metastatic lung cancer diagnosed from 2016 to 2025 at a single institution. Patients without brain metastases at diagnosis formed the training cohort, while patients presenting with synchronous brain metastases formed the testing cohort. Clinical variables and thoracic imaging features from diagnostic chest CT reports were used to train LASSO logistic regression and random forest models, evaluated via AUC and standard diagnostic metrics. Feature contributions were assessed using logistic regression coefficients and random forest permutation importance.
RESULTS: Of 305 patients, 182 were used for training and 123 comprised the testing cohort. Logistic regression achieved AUCs of 0.887, 0.811, and 0.801 for EGFR+, SCLC, and EGFR-, respectively; random forest achieved 0.907, 0.783, and 0.810. Adding imaging features to clinical variables improved prediction for EGFR + and SCLC. Contributory imaging features included fibrosis, emphysema, miliary pattern, cavitation, pleural attachment, and vessel encasement, among others. Three-way prediction showed highest positive predictive value for EGFR + and high negative predictive value for SCLC.
CONCLUSION: Combined clinical and chest imaging features can help predict lung cancer subtype, supporting early multidisciplinary discussions for patients with synchronous brain metastases.