Early Detection of Pancreatic Cancer Using Ensemble Learning with Medical Imaging
Kanchan Yadav, Manish Kumar Sharma, A. Deepak, Kannan Mayuri, Deepika Arora, Pankaj Kumar Singh · 2023
Since pancreatic cancer is frequently discovered at a more advanced stage, there are few effective treatments available, and the prognosis is generally dismal. This work provides a novel method for early detection that combines ensemble learning methods with diagnostic imaging. The research makes use of secondary information from many sources and takes a logical approach that is based on an interpretive mindset. The effectiveness of three combined models - random forest, Progressive Boosting Machines (GBM), as well as AdaBoost - in distinguishing between aggressive and benign ductal lesions is thoroughly assessed in this work. With random forest models achieving an accuracy of 92.5%, results show astounding performance. Specifically in random forest modeling, where the significance of feature scores considerably increases discriminative influence, the influence of feature selection strategies on streamlining input data is underlined. Additionally, the models show strong generalizability across a wide range of datasets, proving their flexibility to actual clinical circumstances. The models' consistency in the face of anticipated data fluctuation is highlighted through robustness analyses, which include sensitivity evaluations and perturbations tests.