OC 20.2 - Artificial intelligence for the point-of-care ultrasound-based detection of joint effusion in patients with hemophilia

Sunday, July 10, 2022

11:00 AM – 11:15 AM

Location: ExCel Center, ICC Capital Suite Room 1

Presentation Format: In-Person

Presenting Author(s)

Roberta Gualtierotti, MD, PhD

Assistant Professor
Fondazione IRCCS Ca' Granda - Ospedale Maggiore Policlinico, Centro Emofilia e Trombosi Angelo Bianchi Bonomi
Milan, Lombardia, Italy

Co-Author(s)

SA

Sara Arcudi, MD

Academic researcher, MD, PhD
Fondazione IRCCS Ca' Granda - Ospedale Maggiore Policlinico, Centro Emofilia e Trombosi Angelo Bianchi Bonomi
Milan, Lombardia, Italy
MC

Marco Colussi

PhD
Department of Computer Science, University of Milan
Milan, Lombardia, Italy
SM

Sergio Mascetti

Associate Professor
Department of Computer Science, University of Milan
Milan, Lombardia, Italy
GC

Gianluca Cafasso

MD
Division of Hematology, University of Milan, Milan, Italy.
Milan, Lombardia, Italy
Eugenia Biguzzi, MD

Hematologist
Hematology clinic, Oncology Institute of Southern Switzerland, EOC
Bellinzona, Ticino, Switzerland
CB

Claudio Bettini

Full Professor
Department of Computer Science, University of Milan
Milan, Lombardia, Italy
Flora Peyvandi, MD, PhD

Professor of Internal Medicine; Director of the Angelo Bianchi Bonomi Hemophilia & Thrombosis Center
Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy; Department of Pathophysiology and Transplantation, Università degli Studi di Milano, Milan, Italy
Milan, Lombardia, Italy

Background: Musculoskeletal ultrasound (MSK-US) is a non-invasive and easily accessible diagnostic tool for joint health assessment of patients with hemophilia. The early identification of hemarthrosis by a remote MSK-US-based system performed by general practitioners or the patients themselves sending images to the Comprehensive Care Center could lead to a better personalized management of patients with hemophilia. A computer-aided diagnosis (CAD) system for the automatic detection of joint effusion could support the physicians in prioritizing interventions

Aims: Our aim was to assess the feasibility of the CAD system by developing a deep-learning algorithm to automatically recognize joint capsule distension in MSK-US images.

Methods: Images of the suprapatellar bursa longitudinal scan (SPB-LS) of the knee at 30°-flexion were collected and labeled by an expert MSK-US operator. The learning algorithm is based on an object detection framework that is trained to detect the normal and the distended joint recesses

Results: We recruited two-hundred consecutive adult patients with hemophilia referring to our Center from October 2020 to December 2021, aged 44.7 ± 18.6 years and 50 sex- and age-matched healthy controls. A total of 8,634 MSK-US images were collected (2,267 knee scans). Of these, 450 were considered valid (SPB-LS) and used for the learning task. We used 330 images for training and 120 images for testing the classifier, leading to the detection of joint recess distensions with an 88% accuracy (80% sensitivity; 93% specificity; Figure 1).

Conclusion(s): Our AI-based algorithm showed that the CAD system for the automatic detection of joint recess distension in patients with hemophilia is feasible. Performance may further improve with a larger training set. This is the first step for the development of a telemedicine system to early detect hemarthrosis in patients with hemophilia, thus prompting early personalized management with the most appropriate therapeutic intervention.