AUTHOR=Chang Jeremy , Liu Yanan , Saey Stephanie A. , Chang Kevin C. , Shrader Hannah R. , Steckly Kelsey L. , Rajput Maheen , Sonka Milan , Chan Carlos H. F. 

TITLE=Machine-learning based investigation of prognostic indicators for oncological outcome of pancreatic ductal adenocarcinoma

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.895515

DOI=10.3389/fonc.2022.895515

ISSN=2234-943X

ABSTRACT=Pancreatic ductal adenocarcinoma (PDAC) is an aggressive malignancy with a poor prognosis. Surgical resection remains the only potential curative treatment option for early-stage resectable PDAC. Patients with locally advanced or micro-metastatic disease should ideally undergo neoadjuvant therapy prior to surgical resection for optimal treatment outcome. Computerized tomography (CT) scan is the most common imaging modality obtained prior to surgery. However, the ability of CT scans to assess nodal status and resectability remains suboptimal and depends heavily on physician experience. Improved preoperative radiographic tumor staging with prediction of postoperative margin and lymph node status could have important implications in treatment sequencing. This paper proposes a novel machine learning predictive model, utilizing a 3-dimensional convoluted neural network (3D-CNN), to reliably predict the presence of lymph node metastasis and postoperative positive margin status based on preoperative CT scans. 881 CT scans were obtained from 110 patients with PDAC. Patients and images were separated into training and validation groups for both lymph node and margin prediction study. Per-scan analysis and per-patient analysis (utilizing majority voting method) were performed. For lymph node prediction 3D-CNN model, accuracy was 90% for per-patient analysis and 75% for per-scan analysis. For postoperative margin prediction 3D-CNN model, accuracy was 81% for per-patient analysis and 76% for per-scan analysis. This paper provides a proof-of-concept that utilizing radiomics and 3D-CNN deep learning framework may be used preoperatively to improve prediction of positive resection margins as well as presence of lymph node metastatic disease. Further investigations should be performed with larger cohorts to increase generalizability of this model; however, there is a great promise in use of CNNs to assist clinicians with treatment selection for patients with PDAC.