Deep learning of bone metastasis in small cell lung cancer: A large sample-based study

Area of Science:

• Oncology
• Medical Informatics

Background:

• Bone metastasis (BM) is common in small cell lung cancer (SCLC) and associated with poor prognosis.
• Predictive models for outcomes in SCLC patients with BM are lacking.

Purpose of the Study:

• To explore epidemiological characteristics of BM in SCLC patients.
• To develop and validate deep learning-based nomograms for predicting cancer-specific survival (CSS) and overall survival (OS) in SCLC patients with BM.

Main Methods:

• Utilized SEER database (2010-2016) for SCLC patient data.
• Employed Cox regression, nomogram construction, and decision tree analysis for prognostic factor identification and model validation.
• Assessed nomogram accuracy using C-index, calibration curves, and time-dependent ROC curves.

Main Results:

• Identified male sex, tumor size, brain/liver metastases, and chemotherapy as risk factors for BM.
• Established prognostic variables for OS (including tumor size, age, N stage, gender, liver metastasis, radiotherapy, chemotherapy) and CSS (adding tumor number).
• Validated nomograms demonstrated good predictive accuracy and clinical utility.

Conclusions:

• The developed nomogram and decision tree models effectively guide treatment decisions for SCLC patients with BM.
• Deep learning models show promise in facilitating the creation of prediction models for SCLC patients with BM.