Developing a Stacked Ensemble-Based Classification Scheme to Predict Second Primary Cancers in Head and Neck Cancer Survivors

Abstract

Despite a considerable expansion in the present therapeutic repertoire for other malignancy managements, mortality from head and neck cancer (HNC) has not significantly improved in recent decades. Moreover, the second primary cancer (SPC) diagnoses increased in patients with HNC, but studies providing evidence to support SPCs prediction in HNC are lacking. Several base classifiers are integrated forming an ensemble meta-classifier using a stacked ensemble method to predict SPCs and find out relevant risk features in patients with HNC. The balanced accuracy and area under the curve (AUC) are over 0.761 and 0.847, with an approximately 2% and 3% increase, respectively, compared to the best individual base classifier. Our study found the top six ensemble risk features, such as body mass index, primary site of HNC, clinical nodal (N) status, primary site surgical margins, sex, and pathologic nodal (N) status. This will help clinicians screen HNC survivors before SPCs occur.

Associations between Sex and Risk Factors for Predicting Chronic Kidney Disease

“…The uncertainty of diagnosis usually results from heterogeneity screening and clinical practices; thus, an accurate tool is needed for early prediction to ensure that potential patients receive and comply with preventive health check-ups. Data mining has been successfully used to build predictive models for healthcare prediction tasks [15][16][17][18][19][20][21]. The present study sought to evaluate the novel hypothesis that men and women with CKD possess different risk factors.…”

Diagnosis of Inflammatory Bowel Disease and Colorectal Cancer through Multi-View Stacked Generalization Applied on Gut Microbiome Data

“…Application of stacking on real-world datasets requires making certain choices and tricks which were referred as ’black art’ [ 27 ]. Even when stacking is applied correctly, it might still perform only as well as best base classifier [ 26 ] or with marginal improvement of about 3% compared to the best individual base classifier [ 41 ]. In our study, we achieved results higher than a single base classifier, with improvement of over 8% on test set for inflammatory bowel disease, possibly due to combining prediction probabilities rather than predicted classification labels themselves, which is in line with general recommendations [ 42 ].…”

“…This recommendation is empirical and we cannot guarantee that stacking is going to work better or worse on other diseases or patient cohorts, because there is no consensus on the most optimal stacking configuration [ 27 ]. However, conclusions and examples provided in this subsection suggest that stacking would perform better than a single model on complex biomedical phenomena with multiple processes involved on molecular level, e.g., aging [ 28 ], head and neck cancer [ 41 ], pregnancy [ 43 ] or protein sequence compression [ 47 ]. Such biological processes can have patterns which can be reflected on multiple levels e.g., proteomics, metabolomics, metagenomics, patient history data etc.…”

Comparison between linear regression and four different machine learning methods in selecting risk factors for osteoporosis in a Chinese female aged cohort

“…These methods were selected as they have been used in different healthcare applications and do not require any prior assumptions about data distribution. [19][20][21][22][23][24][25][26][27][28] To evaluate the efficacy of our proposed scheme, we used MLR as a benchmark for comparison. We also identify the importance of various risk factors for predicting T-score.…”