Detection and characterization of pancreatic lesions with 3D CT data using an automated deep learning-based solution.
Authors
Affiliations (26)
Affiliations (26)
- Philips Research France, 92150 Suresnes, France.
- Department of Radiology, Hôpital de la Croix-Rousse, Hospices Civils de Lyon 69004 Lyon, France. Electronic address: [email protected].
- Department of Radiology, University Hospital of Nancy, Laboratoire IADI INSERM U 1254, 54035 Nancy, France.
- Department of Radiology, Hôpital de la Croix-Rousse, Hospices Civils de Lyon 69004 Lyon, France.
- CentraleSupelec, INRIA, CVN, Université Paris-Saclay, 91190 Gif-sur-Yvette, France.
- Société Française de Radiologie, 75013 Paris, France.
- Department of Radiology, Hôpital Paris Saint Joseph, 75014 Paris, France.
- Department of Diagnostic and Interventional Radiology, Assistance Publique-Hopitaux de Paris, CHU de Bicêtre, 94270 Le Kremlin-Bicêtre, France.
- Department of Radiology, HMB, CHU Reims 51100 Reims, France; CReSTIC, Université de Reims-Champagne-Ardenne, UFR Sciences Exactes et Naturelles, 51100 Reims, France.
- Department of Radiology, Hôtel Dieu, CHU Nantes 44093 Nantes, France.
- Department of Radiology, Hôpital Cochin, Assistance Publique-Hopitaux de Paris 75014 Paris, France; Génomique et Signalisation des Tumeurs Endocrines, Institut Cochin, INSERM U 1016, CNRS UMR8104, 75014 Paris, France; Faculté de Médecine, Université Paris Cité, 75006 Paris, France.
- Department of Radiology, Hôpital Paul Brousse, Assistance Publique-Hopitaux de Paris, 94800 Villejuif, France.
- Department of Radiology, Assistance Publique-Hopitaux de Paris, Sorbonne Université, Hôpital Universitaire Pitié-Salpêtrière, 75013 Paris, France.
- Department of Radiology, Centre Hospitalier de Cayenne, Cayenne 97306, France.
- Medical Imaging Department, Lille University Hospital, 59000 Lille, France.
- Department of Radiology, CHU Angers, Angers University Hospital, 49933 Angers, France.
- Department of Radiology, Hospital Lapeyronie, CHU Montpellier 34000 Montpellier, France.
- Department of Radiology, Gustave Roussy, 94805 Villejuif, France.
- Department of Radiology, Normandie Université, UNIROUEN, Quantif-LITIS EA 4108, Rouen University Hospital, 76031 Rouen, France.
- Department of Radiology, Hôpital Beaujon, AP-HP.Nord, 92110 Clichy, France.
- Department of Radiology, University of Bourgogne Franche-Comté, CHU Besançon 25030 Besançon, France.
- Department of Radiology, Hopital Henri Mondor, Assistance Publique-Hopitaux de Paris, University Paris Est Créteil 94000 Créteil, France.
- Société Française de Radiologie, 75013 Paris, France; Department of Radiology, Hopital Henri Mondor, Assistance Publique-Hopitaux de Paris, University Paris Est Créteil 94000 Créteil, France; INSERM, U955, Team 18, 94000 Créteil, France.
- Société Française de Radiologie, 75013 Paris, France; Department of Diagnostic and Interventional Radiology, Assistance Publique-Hopitaux de Paris, CHU de Bicêtre, 94270 Le Kremlin-Bicêtre, France; Laboratoire d'Imagerie Biomédicale Multimodale Paris-Saclay, Inserm, CNRS, CEA, BIOMAPS, UMR 1281, Université Paris-Saclay, Villejuif 94800, France.
- Department of Radiology, Gustave Roussy, 94805 Villejuif, France; Laboratoire d'Imagerie Biomédicale Multimodale Paris-Saclay, Inserm, CNRS, CEA, BIOMAPS, UMR 1281, Université Paris-Saclay, Villejuif 94800, France.
- Department of Radiology, Hôpital de la Croix-Rousse, Hospices Civils de Lyon 69004 Lyon, France; CREATIS, INSA - Lyon, Universite Claude Bernard Lyon 1, UJM-Saint Etienne, CNRS, Inserm, CREATIS UMR 5220, U1294 Lyon, France.
Abstract
Pancreatic ductal adenocarcinoma (PDA) is a leading cause of cancer-related deaths, with early diagnosis hampered by nonspecific symptoms and limitations of existing imaging techniques. This study aimed to develop a deep learning (DL) algorithm to automatically classify pancreas lesions on contrast-enhanced CT scans as normal, benign, or malignant, to assist radiologists in detecting early-stage pancreatic cancer. A dataset of 1,037 portal-phase CT scans was compiled from 18 institutions. The dataset was divided into a training set (N = 732) and a test set (N = 305), which was further divided into an internal validation test set (N = 139) and an external validation test set (N = 166). After segmentation using the TotalSegmentator algorithm, the pancreas was isolated from each CT scan. A DL model combining TotalSegmentator's pre-trained encoder and nnUNet decoder layers was developed to classify pancreas lesions. Ten-fold cross-validation was applied, and model performance was assessed using precision, recall, area under the curve (AUC) and a final score (FS) representing a weighted average of the previous three metrics. The final prediction combined the outputs of ten models. Across all validation datasets (139 and 166 patients, respectively, in the internal and external dataset), precision and recall were 0.57 and 0.63, respectively, while AUC was 0.84. In the external validation dataset, malignant lesions were detected with an AUC of 0.97. The model achieved an FS of 0.72 in both internal and external validation datasets, indicating consistent performance across datasets. This study demonstrated the feasibility of using a DL algorithm for automated pancreas lesion classification in CT scans. The model showed strong performance, particularly in detecting malignant lesions. Further research is needed to assess the model's clinical applicability and performance in real-world settings.