Predicting Response to Chemotherapy in Patients With Newly Diagnosed High-Risk Neuroblastoma: A Report From the International Neuroblastoma Risk Group

Abstract

PURPOSE Metaiodobenzylguanidine (MIBG) scans are a radionucleotide imaging modality that undergo Curie scoring to semiquantitatively assess neuroblastoma burden, which can be used as a marker of therapy response. We hypothesized that a convolutional neural network (CNN) could be developed that uses diagnostic MIBG scans to predict response to induction chemotherapy. METHODS We analyzed MIBG scans housed in the International Neuroblastoma Risk Group Data Commons from patients enrolled in the Children's Oncology Group high-risk neuroblastoma study ANBL12P1. The primary outcome was response to upfront chemotherapy, defined as a Curie score ≤ 2 after four cycles of induction chemotherapy. We derived and validated a CNN using two-dimensional whole-body MIBG scans from diagnosis and evaluated model performance using area under the receiver operating characteristic curve (AUC). We also developed a clinical classification model to predict response on the basis of age, stage, and MYCN amplification. RESULTS Among 103 patients with high-risk neuroblastoma included in the final cohort, 67 (65%) were responders. Performance in predicting response to upfront chemotherapy was equivalent using the CNN and the clinical model. Class-activation heatmaps verified that the CNN used areas of disease within the MIBG scans to make predictions. Furthermore, integrating predictions using a geometric mean approach improved detection of responders to upfront chemotherapy (geometric mean AUC 0.73 v CNN AUC 0.63, P < .05; v clinical model AUC 0.65, P < .05). CONCLUSION We demonstrate feasibility in using machine learning of diagnostic MIBG scans to predict response to induction chemotherapy for patients with high-risk neuroblastoma. We highlight improvements when clinical risk factors are also integrated, laying the foundation for using a multimodal approach to guiding treatment decisions for patients with high-risk neuroblastoma.

Predicting Neuroblastoma Patient Risk Groups, Outcomes, and Treatment Response Using Machine Learning Methods: A Review

“…Predicting response to treatment in NB patients can be of significant clinical value [ 27 ]. A study attempted to use a CNN method to analyse diagnostic metaiodobenzylguanidine (MIBG) scans and thereby predict response to chemotherapy in NB.…”

“…A study attempted to use a CNN method to analyse diagnostic metaiodobenzylguanidine (MIBG) scans and thereby predict response to chemotherapy in NB. Accordingly, MIBG scans utilise radionucleotides to image patients, and this imaging then contributes to the Curie score for the disease burden assessment [ 27 ]. Despite this, no method can currently predict responses to induction chemotherapy and the related Curie score ≤ 2 at the time of diagnosis.…”

“…The patients with a Curie score ≤ 2 responded to four cycles of induction chemotherapy (n = 67), whereas the nonresponders (n = 36) displayed a Curie score >2 and were more likely to have advanced INSS stage of disease and non-MYCN amplification status. As such, 82% of responders were MYCN-amplified, and 56% of non-MYCN-amplified patients were responders [ 27 ].…”

“…RF is an ensemble ML algorithm that can process complex interactions and classification features, learns rapidly, and performs robustly even with missing data [ 26 ]. Moreover, logistic regressions are used to make binary or multi-class variable predictions and are borrowed by ML methods to predict the probabilities of classes [ 27 ]. Deep learning (DL), as a subsection of ML, is generally used synonymously with deep neural networks (DNN) and can transform data through various representation levels.…”

“…DL (DNN) can also predict clinical outcomes in NB patients [ 28 ], and it can be utilised for image detection and analysis in cancer [ 29 ]. Also, convolutional neural networks (CNN), as a type of DL (DNN), have been shown to learn patterns within images for classification and object recognition [ 27 ] ( Figure 1 ). Finally, ML can be supervised or unsupervised; in the supervised fashion, the input data is labelled, while it is not in the unsupervised learning method.…”