P0012 - Deep Learning and Digital Single-Operator Cholangioscopy: Automatic Detection and Classification of Malignant Biliary Masses in Patients With Indeterminate Biliary Strictures
Centro Hospitalar de São João Porto, Porto, Portugal
Award: Presidential Poster Award
Miguel Mascarenhas, MD1, Tiago Ribeiro, MD1, Joao Afonso, MD1, João Ferreira, PhD2, Filipe Vilas Boas, MD1, Hélder Cardoso, MD1, Renato Natal, PhD2, Pedro Pereira, MD1, Guilherme Macedo, MD, PhD, FACG1 1Centro Hospitalar de São João, Porto, Porto, Portugal; 2University of Porto, Porto, Porto, Portugal
Introduction: The characterization of indeterminate biliary strictures is a significant diagnostic challenge. Digital cholangioscopy allows direct visual inspection of the morphology of biliary strictures, which is highly sensitive for the diagnosis of biliary malignancy. Morphologic characteristics such as masses, tumor vessels or papillary projections are common findings in those with malignant biliary strictures. However, the diagnosis by visual impression remains limited by suboptimal specificity and significant interobserver variability. Artificial intelligence (AI) tools have recently been applied with success to a wide spectrum of endoscopic modalities. The application of these systems for automatic characterization of biliary lesions during cholangioscopy has not been explored. This study aimed to develop a neural network for automatic detection of biliary masses in D-SOC images.
Methods: A convolutional neural network (CNN) was designed for automatic identification of biliary masses in D-SOC images. A total of 3580 frames were extracted from a pool of 85 cholangioscopies (Spyglass™ DS II system, Boston Scientific, Marlborough, MA, USA). Each frame was classified by two endoscopists with experience in D-SOC regarding the presence or absence of masses. A diagnosis of a benign biliary stricture was made in the case of negative histopathology of biopsy or surgical specimens and no evidence of malignancy during a 6-month follow-up period.
Results: The overall accuracy of the deep learning system was 99.9%. After the optimization of the architecture of the CNN for automatic detection of biliary masses, our network showed a sensitivity of 100%, a specificity of 99.8%, a PPV of 99.6% and a NPV of 100%. The AUC was 1.00. The image processing speed was of 17 ms/image.
Discussion: Our neural network demonstrated high performance levels for the detection of biliary masses. An accurate automatic detection of this and other macroscopic characteristics associated with biliary malignancy may positively influence the diagnostic capacity of cholangioscopy, which may ultimately translate into prognostic gains for patients with indeterminate biliary strictures.
Figure: Figure 1: 1a - Output obtained from the application of the CNN. A blue bar represents a correct prediction. 1b - Evolution of the accuracy in predicting the diagnosis of masses associated with biliary malignancy by the CNN during training and validation phases, as the training and validation datasets were repeatedly inputted in the neural network. 1c - ROC analyses of the network’s performance. AUC: area under the curve; CNN: convolutional neural network; B: benign finding; SM: masses.
Disclosures:
Miguel Mascarenhas indicated no relevant financial relationships.
Tiago Ribeiro indicated no relevant financial relationships.
Joao Afonso indicated no relevant financial relationships.
João Ferreira indicated no relevant financial relationships.
Filipe Vilas Boas indicated no relevant financial relationships.
Hélder Cardoso indicated no relevant financial relationships.
Renato Natal indicated no relevant financial relationships.
Pedro Pereira indicated no relevant financial relationships.
Guilherme Macedo indicated no relevant financial relationships.
Miguel Mascarenhas, MD1, Tiago Ribeiro, MD1, Joao Afonso, MD1, João Ferreira, PhD2, Filipe Vilas Boas, MD1, Hélder Cardoso, MD1, Renato Natal, PhD2, Pedro Pereira, MD1, Guilherme Macedo, MD, PhD, FACG1. P0012 - Deep Learning and Digital Single-Operator Cholangioscopy: Automatic Detection and Classification of Malignant Biliary Masses in Patients With Indeterminate Biliary Strictures, ACG 2021 Annual Scientific Meeting Abstracts. Las Vegas, Nevada: American College of Gastroenterology.