A Potential Radiomics-Clinical Model for Predicting Failure of Lymph Node Control after Definite Radiotherapy in...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Predicting lymph node (LN) failure is crucial for optimizing radiotherapy (RT) in head and neck cancer (HNC) with LN metastases.
• Radiomics analysis of CT images during RT may offer insights into treatment response and failure prediction.

Purpose of the Study:

• To evaluate radiomics models for predicting LN failure after definitive RT in HNC.
• To compare the predictive performance of models using CT scans from different time points during RT (pre-RT and mid-RT).

Main Methods:

• Retrospective analysis of 66 HNC patients treated with RT (≥60 Gy) from 2010-2021.
• Radiomics features extracted from pre-treatment (CTpre) and mid-treatment (CTmid) simulation CT scans.
• Development and comparison of radiomics-alone and radiomics plus clinical parameter models using LASSO feature selection.

Main Results:

• Radiomics models incorporating clinical parameters showed improved accuracy and AUC compared to radiomics-alone models.
• The model using mid-treatment CT (CTmid) with clinical parameters achieved the highest mean accuracy (0.790) and AUC (0.662).
• Key clinical parameters included smoking status, T-stage, extranodal extension (ECE), and LN regression rate.

Conclusions:

• Adding clinical parameters significantly improved the predictive performance of radiomics models for LN failure.
• The radiomics model utilizing mid-treatment CT scans and clinical data demonstrated superior predictive capability in this preliminary study.
• This approach holds promise for optimizing treatment strategies and predicting outcomes in HNC patients.