Prediction of MYCN Amplification, 1p and 11q Aberrations in Pediatric Neuroblastoma via Pre-therapy 18F-FDG PET/CT Radiomics

Luodan Qian, Yang Shen, Shuxin Zhang, Hong Qin, Wei Wang, Ying Kan, Lei Liu, Jixia Li, Hui Zhang, Jigang YangFrontiers in Medicine 2022 · DOI

TL;DR

This study demonstrates that radiomic features extracted from pre-therapy 18F-FDG PET/CT scans can accurately predict key genetic aberrations (MYCN amplification, 1p and 11q abnormalities) in pediatric neuroblastoma. Combining these radiomic features with clinical data resulted in the most robust predictive models, offering a non-invasive tool to aid in risk stratification and personalized treatment for this challenging childhood cancer.

What Problem Does This Paper Address?

Neuroblastoma (NB) is a common and often aggressive childhood cancer. Accurate risk stratification is crucial for tailoring treatment and improving patient outcomes. Traditionally, this involves assessing clinical factors, tumor stage, and genetic markers like MYCN amplification and chromosome aberrations (1p and 11q loss). While these genetic markers are powerful prognostic indicators, their assessment often requires invasive procedures like biopsies. Furthermore, while 18F-FDG PET/CT is widely used for NB diagnosis and staging, its potential to predict underlying tumor biology non-invasively has been underexplored. This paper aims to bridge this gap by investigating whether radiomic analysis of pre-treatment 18F-FDG PET/CT scans can predict these critical genetic abnormalities, offering a less invasive alternative for risk assessment.

Methods

The researchers retrospectively analyzed data from 122 pediatric patients diagnosed with neuroblastoma. They developed three types of predictive models:

  1. Clinical Model (C-model): This model used established clinical characteristics (e.g., LDH, NSE, SF, age, MTD) identified as significant predictors through logistic regression.
  2. Radiomic Model (R-model): This model leveraged radiomic features extracted from 18F-FDG PET/CT images. They used the pyradiomics package to extract a wide range of features (first-order, shape, texture matrices like GLCM, GLRLM, GLSZM, NGTDM, GLDM) from both CT and PET components of the scans. The Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed to select the most relevant radiomic features and build predictive models.
  3. Combined Model (CR-model): This model integrated both clinical characteristics and selected radiomic features to potentially enhance predictive power.

The performance of these models was evaluated using Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) metrics on both a training set (n=86) and a separate test set (n=36). Nomograms were also developed to visualize the predictive probability based on the models.

Datasets

The study utilized a retrospective dataset of 122 pediatric patients with newly diagnosed neuroblastoma from a single center (Beijing Friendship Hospital). The patients were diagnosed between March 2018 and November 2019. The dataset was divided into a training set (70% of patients, n=86) and a test set (30% of patients, n=36) for model development and validation.

Key Insights and Ideas

  • Radiomics as a Non-Invasive Biomarker: The study strongly suggests that radiomic features from standard 18F-FDG PET/CT scans can serve as powerful non-invasive surrogates for genetic markers like MYCN amplification and 1p/11q aberrations. This is a significant advancement as it could reduce the need for invasive biopsies for initial risk stratification.
  • Synergy of Modalities: Combining radiomic features with clinical data (CR-model) consistently outperformed models relying solely on clinical or radiomic features. This highlights the complementary nature of these data sources and the potential for a more comprehensive understanding of tumor biology.
  • Wavelet Features Shine: A notable finding is that many of the selected radiomic features were derived from wavelet decomposition of the images. This suggests that these transformed features capture complex, high-order textural information that is highly predictive of the underlying genetic status, information that might be missed by standard visual assessment.
  • Clinical Utility for Risk Stratification: The developed models, particularly the CR-models, demonstrated high predictive accuracy, indicating their potential to directly assist clinicians in stratifying patients into different risk groups, guiding treatment decisions, and monitoring treatment response.

Key Results

  • High Predictive Performance: The radiomic models (R-models) showed strong performance in predicting MYCN, 1p, and 11q status, with AUCs ranging from 0.84 to 0.96 across training and test sets.
  • Superior Combined Models: The combined models (CR-models) achieved the highest predictive accuracy, with AUCs reaching up to 0.98 in the training set and 0.96 in the test set for MYCN prediction. For 1p and 11q, CR-models achieved AUCs of 0.91 and 0.93 (training) and 0.88 and 0.89 (test), respectively.
  • Clinical Predictors Identified: Specific clinical variables like LDH, NSE, SF, and MTD were found to be significant predictors in the C-models for different genetic aberrations.
  • Nomograms for Practical Use: Nomograms were developed, providing a visual tool for estimating the probability of MYCN, 1p, and 11q status based on rad-scores and clinical parameters, facilitating clinical application.

Limitations and Future Work

The authors acknowledge several limitations:

  • Single-Center Study: The data was collected from a single center, which may limit the generalizability of the findings to other patient populations and healthcare settings.
  • Small Cohort Size: While 122 patients is a reasonable number, a larger, multi-center cohort would be beneficial for validating the models and improving their robustness and specificity.
  • Retrospective Design: The study was retrospective, which inherently has limitations compared to prospective studies.

Future work should focus on:

  • Multi-center Validation: Conducting prospective, multi-center studies to validate these radiomic models and confirm their clinical utility.
  • Expanding Radiomic Features: Exploring a broader range of radiomic features and advanced machine learning techniques to further enhance predictive accuracy.
  • Integration with Other Imaging Modalities: Investigating the integration of radiomics from other imaging modalities (e.g., MRI) with PET/CT for a more comprehensive predictive framework.
  • Clinical Implementation Studies: Designing studies to directly assess the impact of these radiomic-based predictions on clinical decision-making and patient outcomes.

Summary updated

Abstract

Purpose This study aimed to assess the predictive ability of 18F-FDG PET/CT radiomic features for MYCN, 1p and 11q abnormalities in NB. Method One hundred and twenty-two pediatric patients (median age 3. 2 years, range, 0.2–9.8 years) with NB were retrospectively enrolled. Significant features by multivariable logistic regression were retained to establish a clinical model (C_model), which included clinical characteristics. 18F-FDG PET/CT radiomic features were extracted by Computational Environment for Radiological Research. The least absolute shrinkage and selection operator (LASSO) regression was used to select radiomic features and build models (R-model). The predictive performance of models constructed by clinical characteristic (C_model), radiomic signature (R_model), and their combinations (CR_model) were compared using receiver operating curves (ROCs). Nomograms based on the radiomic score (rad-score) and clinical parameters were developed. Results The patients were classified into a training set ( n = 86) and a test set ( n = 36). Accordingly, 6, 8, and 7 radiomic features were selected to establish R_models for predicting MYCN, 1p and 11q status. The R_models showed a strong power for identifying these aberrations, with area under ROC curves (AUCs) of 0.96, 0.89, and 0.89 in the training set and 0.92, 0.85, and 0.84 in the test set. When combining clinical characteristics and radiomic signature, the AUCs increased to 0.98, 0.91, and 0.93 in the training set and 0.96, 0.88, and 0.89 in the test set. The CR_models had the greatest performance for MYCN, 1p and 11q predictions ( P < 0.05). Conclusions The pre-therapy 18F-FDG PET/CT radiomics is able to predict MYCN amplification and 1p and 11 aberrations in pediatric NB, thus aiding tumor stage, risk stratification and disease management in the clinical practice.
BibTeX
@inproceedings{qian2022prediction,
  author = {Luodan Qian and Yang Shen and Shuxin Zhang and Hong Qin and Wei Wang and Ying Kan and Lei Liu and Jixia Li and Hui Zhang and Jigang Yang},
  title = {{Prediction of MYCN Amplification, 1p and 11q Aberrations in Pediatric Neuroblastoma via Pre-therapy 18F-FDG PET/CT Radiomics}},
  booktitle = {Frontiers in Medicine},
  year = {2022},
  doi = {10.3389/fmed.2022.840777},
}
Figure