Enhanced deep learning model for precise nodule localization and recurrence risk prediction following curative-intent surgery for lung cancer

Jihwan Park ¹, Mi Jung Rho ², Mi Hyoung Moon ³

⁰College of Liberal Arts, Dankook University, Cheonan-si, Chungcheongnam-do, Republic of Korea.

Plos One +

|

July 12, 2024

View abstract on PubMed

Summary

This study uses deep learning on CT scans to predict lung cancer recurrence sites. The model helps identify high-risk patients for personalized treatment strategies.

Area Of Science

Radiology
Artificial Intelligence
Oncology

Background

Radical surgery is standard for early-stage lung cancer, but recurrence is common.
Preoperative risk stratification can guide treatment and surveillance.
Computed tomography (CT) imaging offers potential for predicting recurrence.

Purpose Of The Study

To analyze lung cancer sites in CT images using deep learning.
To predict potential recurrence in high-risk lung cancer patients.
To aid in early detection and diagnosis of lung cancer recurrence.

Main Methods

Utilized a Mask Region-based Convolutional Neural Network (MRCNN) deep learning model.
Applied the model to anonymized CT images and clinical data of non-small cell lung cancer patients.
Optimized model performance through preprocessing, dynamic learning rates, and hyperparameter tuning.

Main Results

The deep learning model successfully identified lung cancer recurrence sites on CT images.
Performance metrics included bounding box (0.390), classification (0.034), and mask (0.266) accuracy.
Identified recurrence locations were accurately mapped to highlight areas of concern.

Conclusions

The trained model assists clinicians in focusing on high-risk lung regions for recurrence.
This AI tool acts as a clinical decision support system for lung cancer management.
It aids in the detection and diagnosis of lung cancer, improving patient care.