Please note that the copy function is not enabled for this field.
If you wish to
modify
existing outcomes, please copy and paste the current outcome text into the Update field.
LOGIN
CREATE ACCOUNT
LOGIN
CREATE ACCOUNT
MY TRIALS
REGISTER TRIAL
FAQs
HINTS AND TIPS
DEFINITIONS
Trial Review
The safety and scientific validity of this study is the responsibility of the study sponsor and investigators. Listing a study does not mean it has been endorsed by the ANZCTR. Before participating in a study, talk to your health care provider and refer to this
information for consumers
Download to PDF
Trial registered on ANZCTR
Registration number
ACTRN12625000425493
Ethics application status
Approved
Date submitted
20/02/2025
Date registered
8/05/2025
Date last updated
8/05/2025
Date data sharing statement initially provided
8/05/2025
Type of registration
Prospectively registered
Titles & IDs
Public title
Can use of Artificial Intelligence improve Osteoporosis Detection and Bone Health Assessment?
Query!
Scientific title
Evaluation of an artificial intelligence system for opportunistic detection of osteoporosis on chest radiographs.
Query!
Secondary ID [1]
313540
0
Nil
Query!
Universal Trial Number (UTN)
Query!
Trial acronym
AI For Osteo
Query!
Linked study record
Query!
Health condition
Health condition(s) or problem(s) studied:
osteoporosis
336016
0
Query!
Condition category
Condition code
Musculoskeletal
333380
333380
0
0
Query!
Osteoporosis
Query!
Injuries and Accidents
332592
332592
0
0
Query!
Fractures
Query!
Intervention/exposure
Study type
Interventional
Query!
Description of intervention(s) / exposure
This study is about using Artificial Intelligence (AI) to predict bone health from chest X-rays. Normally, osteoporosis is diagnosed using a DEXA scan, which measures bone mineral density (BMD). However, DEXA scans are not always available to everyone, so researchers are testing whether an AI model can estimate bone density just from a chest X-ray.
Researchers will use Deep Learning AI model designed for image classification. The AI model will be trained using ~80,000 X-rays and ~15,000 DEXA scans from real patients to find patterns in X-rays that relate to bone strength. The model was tested on a retrospective subset of X-rays that were not used during training with 80% sensitivity and 80% specificity for diagnosis of osteoporosis. The goal is to see how close the AI's predictions are to the real bone density measurements.
The chest x-ray and BMD scans will be delivered by accredited radiographers. BMD scan radiation doses are significantly less than a typical chest x-ray and around 1% of the radiation dose of a chest x-ray.
The processing of the deep learning model could be done on-premises (local-server within the hospital) or cloud-based AI system.
Correlating BMD scan to the chest x-ray will only be done once and will be within a period of 3 months.
The setting for this intervention will occur within accredited Radiology practices with Flinders University staff ensuring that BMD appointments occurs within the stipulated 3 months as defined in the protocol.
The prospective portion of this study will evaluate this deep learning AI system for opportunistic detection of osteoporosis on chest radiographs in a community setting. Patients referred to a radiology clinic for a chest-x ray will be screened for eligibility. At regular intervals, representatives from the clinic will provide the investigator and delegated staff with a listing of patients who have had chest x-rays performed within the last 2 weeks. Patients will then be contacted by the research team via phone and assessed if they are eligible and willing to participate in the study. Patient consent will be obtained at the time of performing the DEXA scan. Patients flagged for a follow-up DEXA study will be contacted by designated research staff and referred to an external imaging provider to have the study performed.
Patients attending a radiology clinic for a chest x-ray study will be provided with a short pamphlet or brochure by the attending health professional or clerical staff that outlines the nature of the study and contact details of the research team. Patients would contact the research team to express their interest and would be assessed if they are eligible to participate in the study. Patients flagged for a follow-up DEXA study will be contacted by designated research staff and referred to an external imaging provider to have the study performed.
Recruitment will continue until approximately 1307 DEXA BMD studies have been performed.
Query!
Intervention code [1]
330131
0
Early detection / Screening
Query!
Comparator / control treatment
No control group. There are no split arms and every patient in the study gets a DEXA scan. The AI is not replacing standard of care; it is just being tested alongside it. The aim is to evaluate how close the AI model can get to the DEXA results.
Dual-Energy X-ray Absorptiometry (DEXA) is a widely accepted gold standard for assessing bone mineral density (BMD) and body composition. It provides precise measurements of bone health, which is crucial for diagnosing osteoporosis and osteoarthritis-related bone changes.
Query!
Control group
Uncontrolled
Query!
Outcomes
Primary outcome [1]
340970
0
This is a composite primary outcome. Diagnostic accuracy of the AI model in identifying osteoporosis from chest x-ray images, measured by sensitivity, specificity and area under the curve (AUC).
Query!
Assessment method [1]
340970
0
The AI model’s diagnostic performance will be evaluated by comparing its results against the reference standard of DEXA scan findings. Sensitivity (percentage of concordance between patients with disease as diagnosed with DEXA that are also diagnosed with the AI model), and specificity (percentage of concordance between patients who are normal that are also diagnosed as normal on the AI model) will be calculated, along with AUC to access overall accuracy. The aim will be to achieve 80% sensitivity, 80% specificity and 90% AUC.
Query!
Timepoint [1]
340970
0
The primary analysis will be conducted at the time of the initial chest X-ray interpretation, compared to the corresponding DEXA scan results obtained within 3 months of chest X-ray.
Query!
Secondary outcome [1]
446945
0
Evaluation of ability to predict BMD score based off chest x-ray.
Query!
Assessment method [1]
446945
0
Comparison between BMD score estimated by chest x-ray and actual score from BMD.
Query!
Timepoint [1]
446945
0
BMD study will be performed within 3 months of initial chest X-ray.
Query!
Eligibility
Key inclusion criteria
Inclusion criteria for prospective participants.
Patient is 50 years of age or older
Patient has independent mobility and is capable of transferring to a DEXA scanner table
Patient is willing and able to provide informed consent
Patient has undergone a chest X-ray with both frontal and lateral views
Patient is willing and able to undergo a subsequent DEXA BMD scan within 6 weeks of their initial chest x-ray
Patient must weigh 227kg or less
Query!
Minimum age
50
Years
Query!
Query!
Maximum age
No limit
Query!
Query!
Sex
Both males and females
Query!
Can healthy volunteers participate?
Yes
Query!
Key exclusion criteria
Patient has implants, hardware, foreign material or other devices in the lumbar spine or hips
Patient has severe degenerative changes or a fracture deformity in the lumbar spine or hips
Patient is pregnant
Patient has had a previous radiological or nuclear medicine investigation in the 7 days prior to the planned DEXA scan
Any other condition that prevents the proper positioning of the patient to be able to obtain accurate BMD values.
Patient is unable to remain motionless for the duration of the scan
Query!
Study design
Purpose of the study
Prevention
Query!
Allocation to intervention
Non-randomised trial
Query!
Procedure for enrolling a subject and allocating the treatment (allocation concealment procedures)
Query!
Methods used to generate the sequence in which subjects will be randomised (sequence generation)
Query!
Masking / blinding
Open (masking not used)
Query!
Who is / are masked / blinded?
Query!
Query!
Query!
Query!
Intervention assignment
Single group
Query!
Other design features
Query!
Phase
Not Applicable
Query!
Type of endpoint/s
Efficacy
Query!
Statistical methods / analysis
Prospective Data
Once patient recruitment is completed and all DEXA studies have been performed, the lumbar spine T scores will be extracted and compared against the predicted T scores. Analysis for the detection of osteoporosis will be evaluated by thresholding the predicted T score at different values, calculating the TP, FN, FP and TN values, then calculating sensitivity, specificity, positive predictive value (PPV), and negative predict value (NPV). We will also perform receiver operator characteristic (ROC) curve analysis by plotting the true positive rate against false positive rate and calculating the area under the curve (AUC) to obtain a single overall metric of model performance. Confidence intervals for the diagnostic performance metrics will be estimated by bootstrap Monte Carlo simulations.
Characteristics of recruited participants will be analysed by descriptive statistics and graphical methods. To detect potential biases, subgroup analyses will be performed by participant gender, age, and imaging technique (frontal vs lateral projection), and clinic where DEXA study is performed. Further analyses will also be performed by osteoporosis history.
Query!
Recruitment
Recruitment status
Not yet recruiting
Query!
Date of first participant enrolment
Anticipated
15/05/2025
Query!
Actual
Query!
Date of last participant enrolment
Anticipated
28/08/2026
Query!
Actual
Query!
Date of last data collection
Anticipated
Query!
Actual
Query!
Sample size
Target
1307
Query!
Accrual to date
Query!
Final
Query!
Recruitment in Australia
Recruitment state(s)
SA
Query!
Funding & Sponsors
Funding source category [1]
317997
0
Charities/Societies/Foundations
Query!
Name [1]
317997
0
Bone Health Foundation
Query!
Address [1]
317997
0
Query!
Country [1]
317997
0
Australia
Query!
Primary sponsor type
Charities/Societies/Foundations
Query!
Name
Bone Health Foundation
Query!
Address
Query!
Country
Australia
Query!
Secondary sponsor category [1]
320342
0
University
Query!
Name [1]
320342
0
Flinders University
Query!
Address [1]
320342
0
Query!
Country [1]
320342
0
Australia
Query!
Ethics approval
Ethics application status
Approved
Query!
Ethics committee name [1]
316662
0
Bellberry Human Research Ethics Committee F
Query!
Ethics committee address [1]
316662
0
https://bellberry.com.au/
Query!
Ethics committee country [1]
316662
0
Australia
Query!
Date submitted for ethics approval [1]
316662
0
Query!
Approval date [1]
316662
0
24/10/2024
Query!
Ethics approval number [1]
316662
0
2023-04-449-A-4
Query!
Summary
Brief summary
The prospective portion of this study will evaluate this deep learning AI system for opportunistic detection of osteoporosis on chest radiographs in a community setting. This study will recruit participants 50 years of age or over, who have undergone a chest x-ray in the preceding 6 weeks. Osteoporosis affects one million Australians and is associated with an increased risk of minimal trauma (“fragility”) fractures. There are effective medications for treating osteoporosis, and timely intervention can reduce the risk of future fractures by up to 70% and mortality by 11%. However, a significant treatment gap in osteoporosis exists, and the majority of patients that present to hospital with a minimal trauma fracture are neither assessed nor appropriately managed for osteoporosis. The gold-standard for diagnosing osteoporosis is by measuring bone mineral density (BMD) using dual energy X-ray absorptiometry (DEXA) or bone densitometry. Access to DEXA scanners depends on the limited availability of equipment and Medicare rebate restrictions. Therefore, opportunistic screening for osteoporosis using Artificial Intelligence (AI) technology represents an approach for identifying patients at higher risk of osteoporosis and more likely to benefit from having a DEXA BMD study. The retrospective portion of this study will continue the development of AI technology for analysing chest X-rays to estimate BMD. The prospective portion of this study will evaluate this deep learning AI system for opportunistic detection of osteoporosis on chest radiographs in a community setting. This study will recruit participants 50 years of age or over, who have undergone a chest x-ray in the preceding 6 weeks.
Query!
Trial website
Nil
Query!
Trial related presentations / publications
Query!
Public notes
Query!
Contacts
Principal investigator
Name
138578
0
Dr Minh-Son To
Query!
Address
138578
0
Addend AI Pty Ltd PO Box112, Walkerville SA 5081
Query!
Country
138578
0
Australia
Query!
Phone
138578
0
+61 402649208
Query!
Fax
138578
0
Query!
Email
138578
0
[email protected]
Query!
Contact person for public queries
Name
138579
0
Chee Chong
Query!
Address
138579
0
Addend AI Pty Ltd PO Box112, Walkerville SA 5081
Query!
Country
138579
0
Australia
Query!
Phone
138579
0
+61 402649208
Query!
Fax
138579
0
Query!
Email
138579
0
[email protected]
Query!
Contact person for scientific queries
Name
138580
0
Minh-Son To
Query!
Address
138580
0
Addend AI Pty Ltd PO Box112, Walkerville SA 5081
Query!
Country
138580
0
Australia
Query!
Phone
138580
0
+61 402649208
Query!
Fax
138580
0
Query!
Email
138580
0
[email protected]
Query!
Data sharing statement
Will the study consider sharing individual participant data?
Yes
Will there be any conditions when requesting access to individual participant data?
Persons/groups eligible to request access:
•
Researchers who submit a scientifically feasible and ethically sound proposal
Conditions for requesting access:
•
-
What individual participant data might be shared?
•
Aggregate data and de-identified individual participant data (subject to approval).
What types of analyses could be done with individual participant data?
•
Statistical analysis or other secondary research question deemed feasible and ethically sound in the judgement of the statisticians, data manager and researcher.
When can requests for individual participant data be made (start and end dates)?
From:
After publication of study outcomes
To:
No end date
Where can requests to access individual participant data be made, or data be obtained directly?
•
Interested parties may submit a request to researcher for consideration by emailing
[email protected]
.
Are there extra considerations when requesting access to individual participant data?
No
What supporting documents are/will be available?
No Supporting Document Provided
Results publications and other study-related documents
Documents added manually
No documents have been uploaded by study researchers.
Documents added automatically
No additional documents have been identified.
Download to PDF