Enhancing Diagnostic Accuracy in Fracture Identification on Musculoskeletal Radiographs Using Deep Learning

A Retrospective Multi-reader Study of Diagnostic Performance: Carebot AI Bones 1.2 (Deep Learning Algorithms v1.0), Frýdek-Místek Hospital

This retrospective study aims to evaluate the effectiveness of artificial intelligence (AI) in identifying fractures on musculoskeletal X-rays. By comparing the performance of a deep learning AI model with that of experienced radiologists, we seek to understand how AI can help improve fracture detection accuracy in clinical settings. The study analyzed 600 X-rays from both pediatric and adult patients, focusing on identifying fractures across different body parts, including the foot, ankle, knee, hand, wrist, and more. The findings show that integrating AI can increase radiologists' sensitivity in detecting fractures, potentially improving patient outcomes by reducing the number of missed injuries.

Study Overview

• Status: Completed
• Conditions: Fractures; Musculoskeletal
• Intervention / Treatment: Diagnostic test: Carebot AI Bones

• Study Type: Observational
• Enrollment (Actual): 600

Contacts and Locations

Study Locations

• Czechia
  • Moravskoslezský kraj
    • Frýdek-Místek, Moravskoslezský kraj, Czechia, 73801
      • Nemocnice ve Frýdku-Místku, p.o.

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

The study includes a retrospective cohort of pediatric and adult patients who underwent musculoskeletal radiographs between March 20 and May 8, 2023, in a single-center hospital setting.

Description

Inclusion Criteria:

• Patients aged 1 year or older.
• Musculoskeletal X-rays available in Digital Imaging and Communications in Medicine (DICOM) format.
• At least one digital plain radiograph of an appendicular body part, including the foot, ankle, knee, hand, wrist, elbow, shoulder, or pelvis.

Exclusion Criteria:

• Poor radiographic quality that precludes human interpretation.
• Radiographs of the lumbar, thoracic, and cervical spine, or facial/nasal bones.
• Radiographs that do not meet the inclusion criteria for appendicular body parts.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort

Intervention / Treatment

Radiographs Analyzed Using AI and Radiologist Review; This cohort consists of 600 radiographs collected from pediatric and adult patients, aged 1 to 99 years, who underwent X-ray imaging for musculoskeletal conditions. The radiographs include various body parts such as the foot, ankle, knee, hand, wrist, elbow, shoulder, and pelvis. Fractures were present in 95 cases, while 453 cases showed no fractures.

Diagnostic test: Carebot AI Bones; The use of a deep learning-based artificial intelligence software, Carebot AI Bones version 1.2.2, designed to aid in the detection of fractures on musculoskeletal radiographs. The AI model analyzes digital X-ray images to identify fractures, highlighting areas of interest with bounding boxes.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity of AI Model Compared to Radiologists in Fracture Detection on Musculoskeletal X-rays	This outcome measures the sensitivity of the AI model (Carebot AI Bones 1.2.2) in detecting fractures on musculoskeletal X-rays, compared to the sensitivity of radiologists with varying levels of experience. Sensitivity is calculated as the proportion of true positive fracture cases identified by the AI model and radiologists out of all confirmed fracture cases.	From March 2023 to May 2023 (Retrospective analysis period)

Collaborators and Investigators

• Sponsor: Carebot s.r.o.

Study record dates

Study Major Dates

• Study Start (Actual): March 20, 2023
• Primary Completion (Actual): July 15, 2024
• Study Completion (Actual): July 15, 2024

Study Registration Dates

• First Submitted: October 14, 2024
• First Submitted That Met QC Criteria: October 14, 2024
• First Posted (Actual): October 16, 2024

Study Record Updates

• Last Update Posted (Actual): March 18, 2026
• Last Update Submitted That Met QC Criteria: March 16, 2026
• Last Verified: March 1, 2026

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Wounds and Injuries; Fractures, Bone
• Other Study ID Numbers: CB-BONES-01-FM

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO
• IPD Plan Description: Due to privacy concerns and the retrospective nature of the study, individual participant data (IPD) will not be shared. Data collected contains sensitive medical information that is protected under confidentiality agreements and GDPR regulations.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No