Pathological Response Prediction To Neo-Adjuvant Chemoradiation In Esophageal Carcinoma Using Artificial Intelligence And Radiomics: An Exploratory Analysis

Chufal, K. S.; Chowdhary, R. L.

Abstract

Purpose/Objective(s)
To utilize Artificial Neural Networks (ANN) to predict pathological response after neoadjuvant chemoradiation (NACRT), based on Radiomics data extracted from pre-NACRT Computed Tomography (CT) datasets.

Materials/Methods
Between 2013 to 2019, 201 patients with Stage II-IVa esophageal carcinoma underwent NACRT followed by radical esophagectomy at our institution, out of which 97 patients were randomly selected to form our study cohort. All patients received radiotherapy [median dose: 41.4 Gy (Range: 39.6-50.4) / 23 Fx (Range: 22-28)] via IMRT or 3DCRT technique and concurrent weekly platinum & taxane-based chemotherapy [Median cycles: 5 (Range: 1-6)]. 55 patients achieved a pathological complete response (pCR).

Results
A total of 254 features were extracted from each patient’s CT dataset. RF yielded 15 features with ≥ 95% probability of predicting pathological outcome, and following multivariable logistic regression, 7 features served as the input layer for the MLP model. The selected features described the sphericity and 3-dimensional higher-order features associated with the tumor. The overall accuracy of the model was 80% and 77.8% in the training and validation cohort, respectively (AUC = 0.87).

Conclusion
ANN-based predictive modelling of pathological outcome after NACRT for esophageal carcinomas, utilizing only Radiomic features (after appropriate dimensionality reduction) is feasible and warrants further investigation

Citation

Chufal, K. S., Ahmad, I., Bajpai, R., Miller, A., Chowdhary, R. L., & Gairola, M. (2020, October). Pathological Response Prediction To Neo-Adjuvant Chemoradiation In Esophageal Carcinoma Using Artificial Intelligence And Radiomics: An Exploratory Analysis. Poster presented at Global Oncology: The ASTRO Annual Meeting

• Presentation Conference Type: Poster
• Conference Name: Global Oncology: The ASTRO Annual Meeting
• Start Date: Oct 24, 2020
• End Date: Oct 28, 2020
• Deposit Date: Jun 19, 2023
• Publisher: Elsevier
• DOI: https://doi.org/10.1016/j.ijrobp.2020.07.1860
• Keywords: Cancer Research; Radiology, Nuclear Medicine and imaging; Oncology; Radiation
• Public URL: https://keele-repository.worktribe.com/output/488861
• Additional Information: This article is maintained by: Elsevier; Article Title: Pathological Response Prediction To Neo-Adjuvant Chemoradiation In Esophageal Carcinoma Using Artificial Intelligence And Radiomics: An Exploratory Analysis; Journal Title: International Journal of Radiation Oncology*Biology*Physics; CrossRef DOI link to publisher maintained version: https://doi.org/10.1016/j.ijrobp.2020.07.1860; Content Type: simple-article; Copyright: Copyright © 2020 Published by Elsevier Inc.