Abstract Title:
Predicting 30-Day Outcomes in Head and Neck Cancer Surgery Using Machine Learning

Background:
Readmission following surgical resection of head and neck squamous cell carcinoma (HNSCC) has been reported to range from 7.7% to 26.5%, with worse outcomes and cost increase of $15,123 for care of patients with unplanned readmission. While previous studies have shown correlations between various clinical factors and hospital readmission risk, they have not provided actionable clinical predictions leveraging these factors to impact clinical decisions. This study aims to fill this gap by developing and validating a machine learning model to predict 30-day mortality and hospital readmission likelihood following definitive surgical resection for HNSCC using a large-scale, well-curated dataset. This model may serve as an actionable, transparent, and highly accurate clinical decision support tool to identify high-risk patients and redirect preventative resources effectively.

Methods:
This prognostic predictive modeling study utilized retrospective data from the 2006 to 2018 National Cancer Database (NCDB), including 103,891 patients with HNSCC undergoing surgical resection. Predictors included comprehensive pre-treatment demographic and clinical variables available in the NCDB. A combination of five machine learning models was trained on 80% of the data and tested on the remaining 20% to predict the primary outcomes of 30-day mortality and 30-day readmission.

Results:
Among the 103,891 patients satisfying the inclusion criteria, 5,838 (5.6%) were readmitted, and 829 (0.8%) died within 30 days. The models achieved an AUC of 0.80 (95% CI: 0.77-0.83) for mortality prediction and an AUC of 0.66 (95% CI: 0.65-0.68) for readmission prediction. SHAP value analysis identified sex and urban-rural index as the most predictive variables for mortality and readmission, respectively.

Conclusion:
The machine learning model demonstrated high sensitivity and specificity in predicting 30-day mortality and readmission risks in a large-scale dataset. This model highlights critical features influencing predictions, providing key clinical insights supporting clinical decision-making. With further validation in prospective datasets, this model can serve as an effective clinical decision support tool, guiding resource allocation and follow-up care for patients undergoing surgical resection of HNSCC. This study presents a significant advancement in large-scale validation of a potential clinical decision support tool that is poised to enhance patient outcomes by providing transparent and actionable predictions on patient outcomes.

Predicting 30-Day Outcomes in Head and Neck Cancer Surgery Using Machine Learning

Background:
Breast cancer is the most commonly diagnosed cancer type, accounting for about 2.3 million new cases and 685,000 deaths worldwide in 2020. In the US, breast cancer screenings are routinely performed in women starting from the age of 40. One major challenge with breast screening is the high rate of false positives, with over 50% of women experiencing at least one false positive result over 10 years of annual screening on plain film mammography, the most common breast cancer screening modality. Deep learning methods have been shown to be effective in reducing false positive rates that are associated with unnecessary costs, medical procedures, and emotional burden on patients, as well as improving the accuracy of cancer detection in screening mammograms — hence, this study investigates the use of deep learning methods for breast cancer detection. Our aim was to investigate and evaluate the performance of transfer learning and deep learning models in breast cancer detection.

Methods:
We trained and evaluated our deep learning model using multi-center breast screening data provided by Kaggle in the RSNA Screening Mammography Breast Cancer Detection competition. Our model uses mammogram images as input to fine-tune a pre-trained ResNet-50 deep-learning model via transfer learning and outputs the probability of cancer in each image. The model performance is evaluated using a held-out test-set (train:validate:test 70:10:20) to compute the sensitivity, specificity, and probabilistic F1-score achieved.

Results:
Overall we achieved a classification sensitivity of 1.00 and a specificity of 1.00 (ROC-AUC=1.00) on our held-out test set, with a probabilistic F1 score of 0.795 on our held-out test set. 

Conclusion:
Our model detected breast cancer with 100% sensitivity and specificity in the held-out test set. Future work may focus on validating algorithm performance in a larger out-of-distribution patient cohort. Application of deep learning models to breast cancer detection in mammograms may lead to improved clinical efficiency, diagnostic accuracy, and reduction in costs associated with the high false positive burden in breast cancer screening.


Daniel Fu

2

SHORT BIO

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES