Eelin Tan1, Khurshid Merchant2, Bhanu Prakash Kn3, Arvind Cs3, Joseph J Zhao4, Seyed Ehsan Saffari5, Poh Hwa Tan6, Phua Hwee Tang6. 1. Department of Diagnostic & Interventional Imaging, KK Womens' and Childrens' Hospital, 100 Bukit Timah Rd, Singapore, 229899, Singapore. eelinnn@gmail.com. 2. Department of Pathology and Laboratory Medicine, KK Womens' and Childrens' Hospital, 100 Bukit Timah Rd, Singapore, 229899, Singapore. 3. Bioinformatics Institute, A*Star, 30 Biopolis Street, #07-01 Matrix, Singapore, 138671, Singapore. 4. Yong Loo Lin School of Medicine, National University of Singapore, 10 Medical Dr, Singapore, 117597, Singapore. 5. Center for Quantitative Medicine, Duke-NUS Graduate Medical School, 8 College Rd, Singapore, 169857, Singapore. 6. Department of Diagnostic & Interventional Imaging, KK Womens' and Childrens' Hospital, 100 Bukit Timah Rd, Singapore, 229899, Singapore.
Abstract
PURPOSE: MYCN onco-gene amplification in neuroblastoma confers patients to the high-risk disease category for which prognosis is poor and more aggressive multimodal treatment is indicated. This retrospective study leverages machine learning techniques to develop a computed tomography (CT)-based model incorporating semantic and non-semantic features for non-invasive prediction of MYCN amplification status in pediatric neuroblastoma. METHODS: From 2009 to 2020, 54 pediatric patients treated for neuroblastoma at a specialized children's hospital with pre-treatment contrast-enhanced CT and MYCN status were identified (training cohort, n = 44; testing cohort, n = 10). Six morphologic features and 107 quantitative gray-level texture radiomics features extracted from manually drawn volume-of-interest were analyzed. Following feature selection and class balancing, the final predictive model was developed with eXtreme Gradient Boosting (XGBoost) algorithm. Accumulated local effects (ALE) plots were used to explore main effects of the predictive features. Tumor texture maps were also generated for visualization of radiomics features. RESULTS: One morphologic and 2 radiomics features were selected for model building. The XGBoost model from the training cohort yielded an area under the receiver operating characteristics curve (AUC-ROC) of 0.930 (95% CI, 0.85-1.00), optimized F1-score of 0.878, and Matthews correlation coefficient (MCC) of 0.773. Evaluation on the testing cohort returned AUC-ROC of 0.880 (95% CI, 0.64-1.00), optimized F1-score of 0.933, and MCC of 0.764. ALE plots and texture maps showed higher "GreyLevelNonUniformity" values, lower "Strength" values, and higher number of image-defined risk factors contribute to higher predicted probability of MYCN amplification. CONCLUSION: The machine learning model reliably classified MYCN amplification in pediatric neuroblastoma and shows potential as a surrogate imaging biomarker.
PURPOSE: MYCN onco-gene amplification in neuroblastoma confers patients to the high-risk disease category for which prognosis is poor and more aggressive multimodal treatment is indicated. This retrospective study leverages machine learning techniques to develop a computed tomography (CT)-based model incorporating semantic and non-semantic features for non-invasive prediction of MYCN amplification status in pediatric neuroblastoma. METHODS: From 2009 to 2020, 54 pediatric patients treated for neuroblastoma at a specialized children's hospital with pre-treatment contrast-enhanced CT and MYCN status were identified (training cohort, n = 44; testing cohort, n = 10). Six morphologic features and 107 quantitative gray-level texture radiomics features extracted from manually drawn volume-of-interest were analyzed. Following feature selection and class balancing, the final predictive model was developed with eXtreme Gradient Boosting (XGBoost) algorithm. Accumulated local effects (ALE) plots were used to explore main effects of the predictive features. Tumor texture maps were also generated for visualization of radiomics features. RESULTS: One morphologic and 2 radiomics features were selected for model building. The XGBoost model from the training cohort yielded an area under the receiver operating characteristics curve (AUC-ROC) of 0.930 (95% CI, 0.85-1.00), optimized F1-score of 0.878, and Matthews correlation coefficient (MCC) of 0.773. Evaluation on the testing cohort returned AUC-ROC of 0.880 (95% CI, 0.64-1.00), optimized F1-score of 0.933, and MCC of 0.764. ALE plots and texture maps showed higher "GreyLevelNonUniformity" values, lower "Strength" values, and higher number of image-defined risk factors contribute to higher predicted probability of MYCN amplification. CONCLUSION: The machine learning model reliably classified MYCN amplification in pediatric neuroblastoma and shows potential as a surrogate imaging biomarker.
Authors: Richard E Overman; Tanvi T Kartal; Aaron J Cunningham; Elizabeth A Fialkowski; Bindi J Naik-Mathuria; Sanjeev A Vasudevan; Marcus M Malek; Ranjeet Kalsi; Hau D Le; Linda Cherney Stafford; Timothy B Lautz; Benjamin T Many; Rachel E Jones; Andreana Bütter; Jacob Davidson; Andrew Williams; Roshni Dasgupta; Jana Lewis; Misty Troutt; Jennifer H Aldrink; Sara A Mansfield; Dave R Lal; Jerry Xiao; Rebecka L Meyers; Scott S Short; Erika A Newman Journal: Pediatr Blood Cancer Date: 2020-02-19 Impact factor: 3.167
Authors: Daniel A Morgenstern; Rochelle Bagatell; Susan L Cohn; Michael D Hogarty; John M Maris; Lucas Moreno; Julie R Park; Andrew D Pearson; Gudrun Schleiermacher; Dominique Valteau-Couanet; Wendy B London; Meredith S Irwin Journal: Pediatr Blood Cancer Date: 2018-11-26 Impact factor: 3.167
Authors: Saif F Hassan; Shawn Mathur; Thomas J Magliaro; Emily L Larimer; Lauren B Ferrell; Sanjeev A Vasudevan; Danielle M Patterson; Chrystal U Louis; Heidi V Russell; Jed G Nuchtern; Eugene S Kim Journal: J Pediatr Surg Date: 2012-06 Impact factor: 2.545
Authors: Frank Berthold; Joachim Boos; Stefan Burdach; Rudolf Erttmann; Günter Henze; Johann Hermann; Thomas Klingebiel; Bernhard Kremens; Freimut H Schilling; Martin Schrappe; Thorsten Simon; Barbara Hero Journal: Lancet Oncol Date: 2005-09 Impact factor: 41.316
Authors: Susan L Cohn; Andrew D J Pearson; Wendy B London; Tom Monclair; Peter F Ambros; Garrett M Brodeur; Andreas Faldum; Barbara Hero; Tomoko Iehara; David Machin; Veronique Mosseri; Thorsten Simon; Alberto Garaventa; Victoria Castel; Katherine K Matthay Journal: J Clin Oncol Date: 2008-12-01 Impact factor: 44.544
Authors: Daniel A Morgenstern; Ulrike Pötschger; Lucas Moreno; Vassilios Papadakis; Cormac Owens; Shifra Ash; Claudia Pasqualini; Roberto Luksch; Alberto Garaventa; Adela Canete; Martin Elliot; Aleksandra Wieczorek; Geneviève Laureys; Per Kogner; Josef Malis; Ellen Ruud; Maja Beck-Popovic; Gudrun Schleiermacher; Dominique Valteau-Couanet; Ruth Ladenstein Journal: Pediatr Blood Cancer Date: 2018-07-17 Impact factor: 3.167