A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy

• Published in: PloS one 2019-10, Vol.14 (10), p.e0218642
• Main Authors: Kaissis, Georgios, Ziegelmayer, Sebastian, Lohöfer, Fabian, Steiger, Katja, Algül, Hana, Muckenhuber, Alexander, Yen, Hsi-Yu, Rummeny, Ernst, Friess, Helmut, Schmid, Roland, Weichert, Wilko, Siveke, Jens T, Braren, Rickmer
• Format: Article
• Language: English
• Subjects: Adenocarcinoma; Adult; Algorithms; Antineoplastic agents; Antineoplastic Combined Chemotherapy Protocols - administration & dosage; Artificial intelligence; Cancer therapies; Cancer treatment; Carcinoma, Pancreatic Ductal - metabolism; Carcinoma, Pancreatic Ductal - mortality; Carcinoma, Pancreatic Ductal - pathology; Carcinoma, Pancreatic Ductal - therapy; Care and treatment; Chemotherapy; Computer and Information Sciences; Data mining; Deoxycytidine - administration & dosage; Deoxycytidine - analogs & derivatives; Disease-Free Survival; Entropy; Feature extraction; Female; Fluorouracil - administration & dosage; Gemcitabine; Humans; Immunohistochemistry; Internal medicine; Internet; Irinotecan - administration & dosage; Keratins, Hair-Specific - metabolism; Keratins, Type II - metabolism; Learning algorithms; Leucovorin - administration & dosage; Machine Learning; Male; Medical prognosis; Medical research; Medicine; Medicine and Health Sciences; Middle Aged; Neoplasm Proteins - metabolism; NMR; Nuclear magnetic resonance; Oxaliplatin - administration & dosage; Pancreas; Pancreatic cancer; Pancreatic Neoplasms - metabolism; Pancreatic Neoplasms - mortality; Pancreatic Neoplasms - pathology; Pancreatic Neoplasms - therapy; Pathology; Patients; Physical Sciences; Predictions; Radiomics; Regression analysis; Regression models; Research and Analysis Methods; Retrospective Studies; Sensitivity and Specificity; Survival; Survival Rate; Tumors
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0218642

Development of a supervised machine-learning model capable of predicting clinically relevant molecular subtypes of pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features. The retrospective observational study assessed 55 surgical PDAC patients. Molecular subtypes were defined by immunohistochemical staining of KRT81. Tumors were manually segmented and 1606 radiomic features were extracted with PyRadiomics. A gradient-boosted-tree algorithm was trained on 70% of the patients (N = 28) and tested on 30% (N = 17) to predict KRT81+ vs. KRT81- tumor subtypes. A gradient-boosted survival regression model was fit to the disease-free and overall survival data. Chemotherapy response and survival were assessed stratified by subtype and radiomic signature. Radiomic feature importance was ranked. The mean±STDEV sensitivity, specificity and ROC-AUC were 0.90±0.07, 0.92±0.11, and 0.93±0.07, respectively. The mean±STDEV concordance indices between the disease-free and overall survival predicted by the model based on the radiomic parameters and actual patient survival were 0.76±0.05 and 0.71±0.06, respectively. Patients with a KRT81+ subtype experienced significantly diminished median overall survival compared to KRT81- patients (7.0 vs. 22.6 months, HR 4.03, log-rank-test P =