Abstract
Identification of nodal metastasis and tumor extranodal extension (ENE) is crucial for head and neck cancer management, but currently only can be diagnosed via postoperative pathology. Pretreatment, radiographic identification of ENE, in particular, has proven extremely difficult for clinicians, but would be greatly influential in guiding patient management. Here, we show that a deep learning convolutional neural network can be trained to identify nodal metastasis and ENE with excellent performance that surpasses what human clinicians have historically achieved. We trained a 3-dimensional convolutional neural network using a dataset of 2,875 CT-segmented lymph node samples with correlating pathology labels, cross-validated and fine-tuned on 124 samples, and conducted testing on a blinded test set of 131 samples. On the blinded test set, the model predicted ENE and nodal metastasis each with area under the receiver operating characteristic curve (AUC) of 0.91 (95%CI: 0.85–0.97). The model has the potential for use as a clinical decision-making tool to help guide head and neck cancer patient management.
Highlights
Patient Cohort Primary Cancer Site Oropharynx Oral Cavity Larynx/Hypopharynx/Nasopharynx Salivary Gland Unknown/Other Clinical T-stage T0 T1 T2 T3 T4 Unknown Clinical N-stage N0 N1 N2 N3 Unknown human papilloma virus (HPV)/p16 Status* Negative Positive Unknown Lymph Node Pathology Negative Nodal Metastasis, extranodal extension (ENE)(−) Node Metastasis, ENE(+)
Following pathologic correlation with CT scans, 653 lymph nodes were segmented in total: 380 negative nodes, 153 nodal metastasis (NM) without ENE, and 120 NM with ENE
Median region of interest (ROI) diameter was significantly greater in NM with ENE (23 mm, range: 10–64 mm), than NM without ENE (16 mm, range: 6–42 mm), or negative nodes (10 mm, range: 4–20 mm) (P < 0.001)
Summary
Patient Cohort Primary Cancer Site Oropharynx Oral Cavity Larynx/Hypopharynx/Nasopharynx Salivary Gland Unknown/Other Clinical T-stage T0 T1 T2 T3 T4 Unknown Clinical N-stage N0 N1 N2 N3 Unknown HPV/p16 Status* Negative Positive Unknown Lymph Node Pathology Negative Nodal Metastasis, ENE(−) Node Metastasis, ENE(+). Studies of patients selected to undergo upfront surgery have shown high use of tri-modality treatment, suggesting that incidental finding of pathologic ENE is not uncommon[12,13,14,15]. Both clinically overt and pathologic ENE have been incorporated into the AJCC 8th edition prognostic staging system for HNSCC16. This highlights the importance of ENE identification in the management of HNSCC and the need to develop better methods to better predict in the pretreatment setting. We sought to develop a deep learning model to detect ENE and NM in HNSCC patients on pretreatment diagnostic CT scans
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call