Summary:

We assessed feasibility and effectiveness of AI-based VF screening in CT, integrated with a local FLS. The system identified VFs in 14% of patients, half previously unrecognized or untreated. This suggests that 2–3 patients with VFs were identified daily at our hospital, highlighting the potential clinical impact of AI-assisted detection.

Purpose:

To evaluate the feasibility and efficacy of integrating an AI algorithm into the radiology workflow for opportunistic vertebral fracture (VF) screening in CT and align it to a local fracture liaison service (FLS).

Methods:

The AI algorithm was integrated into the radiology workflow and applied to all non-skeletal CT scans covering thorax and/or abdomen from patients aged ≥ 50 years over a four-month period at our hospital (catchment area ~ 250,000). Detected VFs were verified by radiologists and subsequently referred to the FLS for further management. A system was established to enable both technical and clinical monitoring.

Results:

The AI setup and workflow were considered feasible and robust, and AI showed a high performance. During the study period, 3971 unique patients (mean age 72 ± 11 years; 51% female) underwent 5147 CT scans. The AI algorithm identified VFs in 566 patients (14%, mean age 78 ± 10; 62% women), all of which were confirmed by radiologist. After clinical triage, 49% were considered in need of further osteoporosis evaluation/treatment, the remainder where either terminally ill/died shortly after CT or were considered correctly handled before.

Conclusion:

AI-based opportunistic screening for VF is feasible and effective in routine clinical practice. Integration of such tools into radiology workflows enhances the detection of at-risk patients and supports timely referral to FLS, potentially reducing the burden of untreated osteoporosis and future fracture risk. In our clinical setting, this meant 2–3 new identified patients every day. These findings support the broader implementation of AI in secondary fracture prevention strategies.

Objective: Vertebral fractures are significantly underdiagnosed in routine clinical practice.Opportunistic screening using CT scans could enhance detection rates. This study aims toimplement an MDR approved AI-screening tool (FLAMINGO, IB Lab) within a clinical FractureLiaison Service (FLS), thereby integrating radiology with clinical care.

Materials and Methods: In this prospective study, the AI-screening tool was incorporatedinto clinical care, linking radiology workflows to a FLS. All patients over 50 years of ageundergoing thoracic and/or abdominal CT scans for non-skeletal reasons (e.g. pneumonia,malignancy, kidney stones) at a medium-sized hospital, serving approximately 250,000inhabitants, were included.

Results: During the first month, a total of 1,249 CT scans (1,127 unique patients) werescreened using the AI algorithm, with 19% flagged as positive (238 CT scans [216 uniquepatients]). Radiologist confirmation indicated that 71% of these were true positives (169 CTscans [148 unique patients]). In the FLS triage, 74 of these 148 patients (50%) were referredfor further osteoporosis investigation and treatment. The remaining patients were alreadyknown and under correct treatment (30%) or were terminally ill or had deceased shortly(during 3 months) after the CT (20%). In total, 6.5% of all patients originally screened werefinally referred for a vertebral fracture that was new or had not been handled correctlybefore.

Conclusion: The implementation of the AI-screening tool (FLAMINGO, IB Lab) within theclinical FLS demonstrated significant potential to increase the detection of vertebral fracturesand to enhance subsequent investigation and treatment of osteoporosis.

Introduction: Low contrast resolution in abdominal computed tomography (CT) may be negativelyaffected by attempts to lower patient doses. Iterative reconstruction (IR) algorithms play a key role inmitigating this problem. The reconstructed slice thickness also influences image quality. The aim was toassess the interaction and influence of patient dose, slice thickness, and IR strength on image quality inabdominal CT.Method: With a simultaneous acquisition, images at 42 and 98 mAs were obtained in 25 patients.Multiplanar images with slice thicknesses of 1, 2, and 3 mm and advanced modeled iterative reconstruction (ADMIRE) strengths of 3 (AD3) and 5 (AD5) were reconstructed. Four radiologists evaluated theimages in a pairwise manner based on five image criteria. Ordinal logistic regression with mixed effectswas used to evaluate the effect of tube load, ADMIRE strength, and slice thickness using the visualgrading regression technique.Results: For all assessed image criteria, the regression analysis showed significantly (p < 0.001) higherimage quality for AD5, but lower for tube load 42 mAs, and slice thicknesses of 1 mm and 2 mm,compared to the reference categories of AD3, 98 mAs, and 3 mm, respectively. AD5 at 2 mm was superiorto AD3 at 3 mm for all image criteria studied. AD5 1 mm produced inferior image quality for liver parenchyma and overall image quality compared to AD3 3 mm. Interobserver agreement (ICC) ranged from0.874 to 0.920.Conclusion: ADMIRE 5 at 2 mm slice thickness may allow for further dose reductions due to its superiority when compared to ADMIRE 3 at 3 mm slice thickness.Implications for practice: Combination of thinner slices and higher ADMIRE strength facilitates imaging atlow dose.

Non-linear properties of iterative reconstruction (IR) algorithms can alter image texture. We evaluated the effect of a model-basedIR algorithm (advanced modelled iterative reconstruction; ADMIRE) and dose on computed tomography thorax image quality.Dual-source scanner data were acquired at 20, 45 and 65 reference mAs in 20 patients. Images reconstructed with filteredback projection (FBP) and ADMIRE Strengths 3–5 were assessed independently by six radiologists and analysed using an ordinallogistic regression model. For all image criteria studied, the effects of tube load 20 mAs and all ADMIRE strengths were significant(p < 0.001) when compared to reference categories 65 mAs and FBP. Increase in tube load from 45 to 65 mAs showed imagequality improvement in three of six criteria. Replacing FBP with ADMIRE significantly improves perceived image quality for allcriteria studied, potentially permitting a dose reduction of almost 70% without loss in image quality

Summary: We assessed the prevalence of hypocalcemia after denosumab injections in a real-world cohort routinely monitored for calcium during up to 7.5 years of treatment. Among 1096 injections in 242 patients, 6.3% resulted in hypocalcemia, and was independent of the injection number. Severe hypocalcemia was rare (1%).

Purpose: To assess the prevalence of and risk factors for hypocalcemia after administration of denosumab in a patient cohort routinely monitored for ionized calcium after each dose.

Methods: In this retrospective observational study, we analyzed denosumab-induced hypocalcemia in a real-world cohort who were routinely followed up with ionized calcium pre- and post-injection (within 31 days after injection) during the period 2011 to 2020.

Results: In total, we included data from 1096 denosumab injections in 242 individuals (1–15 injections per patient). The mean age for the first injection was 74 ± 10 years, and 88% were female. Post-injection hypocalcemia occurred after 6.3% of all injections (4.6% mild, 0.6% moderate, and 1.1% severe) and was independent of the number of injections (rate of hypocalcemia varied from 3–8%). Risk factors for hypocalcemia were male sex, severe renal failure, pre-injection hypocalcemia, hypomagnesemia, hypophosphatemia, and vitamin D insufficiency. Furthermore, older age was not associated with an increased hypocalcemia risk.

Conclusions: Denosumab-induced hypocalcemia is a prevalent adverse event, which occurs independently of the number of injections. However, severe hypocalcemia is a rare occurrence, and severe renal failure and nutritional status appear to be important predictive factors. Magnesium and phosphate might add value in the pre-injection risk assessment; however, this observation needs to be confirmed in larger cohorts.

Objective: Patient education is part of fracture liaison services in many organizations and has been shown to have positive effects on several important patient outcomes. The present study aimed to assess the patients’ preferences when seeking disease specific information on osteoporosis. Methods: Patients with denosumab (Den) or zoledronic acid (ZA) treatment administrated at our endocrinology clinic July 2017 – December 2017 were sent a postal questionnaire with questions on preferences on how to obtain information about osteoporosis. One reminder was sent to nonresponders. Participants were mainly patients followed by primary healthcare, but with help from specialized hospital care to administrate injections/infusions. No osteoporosis school was arranged by the health organization in the catchment area. Results: A total number of 155 patients (84% females, mean age±SD 75±9 y) participated in the study. Dropout rate was 26%. A total of 67% of patients actively searched for disease specific information on osteoporosis. No difference in search behavior was seen regarding type of treatment (Den or ZA) or age. More women than men searched for information as did patients experiencing any adverse event (both p<0.01). The most used source of information was brochure (46%), internet sites from healthcare providers (27%), internet in general (15%) and weekly magazines (8%). In total, 57% of participants stated they would like to attend a school for osteoporosis. Of these 44% preferred general lectures, 39% group education with physical meet-ups, 23% internet (general information) and 13% internet osteoporosis schools. Conclusion: Disease specific information is prompted by a majority of osteoporosis patients. Type of education and sources varies. Internet sources, preferentially from the healthcare organization, was used by a third of patients. The study was done pre-pandemic and it is possible that the use of internet resources might have changed.

Objectives:Fracture liaison services are designed to identify patients needing osteoporosis treatment after a fracture. Some fracture liaison service designs involve a prescreening step, for example, fracture risk assessment tool (FRAX®). Another possible prescreening tools are bone mass density assessment in the acute setting. The aim of this study was to assess the effectiveness of prescreening tools.Methods:In the present prospective cohort study, women aged >55 years with a radius fracture were included. Patients were recruited at the emergency department after experiencing their fracture. All patients performed fracture risk assessment by fracture risk assessment tool, and bone mass density assessment by digital X-ray radiogrammetry and dual-energy X-ray absorptiometry (prescreening steps) as well as full routine evaluation at the osteoporosis unit (endpoint). The main outcome measures were sensitivity, specificity, predictive values, and area under the curve.Results:Forty-one women were recruited (mean age: 70 ± 8 years). Of these, 54% fulfilled the treatment indication criteria of osteoporosis after a full examination. Fracture risk assessment tool without bone mass density (cutoff ⩾ 15%) for prescreening patients had a high sensitivity (90%) but a low area under the curve (0.50) and specificity (16%). The highest area under the curve (0.73) was found prescreening with bone mass density assessment (dual-energy X-ray absorptiometry or digital X-ray radiogrammetry) having a sensitivity of 59%–86% and specificity of 61%–90%.Conclusion:This study, though small, raises questions regarding the effectiveness of using a prescreening step in fracture liaison services for high-risk individuals. In this cohort, FRAX® without bone mass density had a low precision, with a risk of both underestimating and overestimating patients requiring treatment. Bone mass density assessment in the acute setting could improve the precision of prescreening. Further investigations on the effectiveness and health economics of prescreening steps in fracture liaison services are needed.

Background: We aimed to compare two volumetric bone mineral density (vBMD) analysis programs, regarding (I) agreement of vBMD values based on monoand dual-energy computed tomography (MECT and DECT) scans and (II) suitability for analyzing DECT data obtained at different energies. Methods: We retrospectively analyzed two abdominal CT datasets: one performed in a MECT scan (vertebrae L1-L3) and one in a DECT scan (vertebrae L1-L4). Each dataset included different individuals [MECT 15 patients (45 vertebrae) and DECT 12 patients (48 vertebrae), respectively]. vBMD analysis was conducted using Philips IntelliSpace (IP) and Mindways qCT Pro (MW). Regarding the DECT scans, vBMD analysis was done at three different energies: 80, 150 and synthetic 120 kVp and for MECT scan at 120 kVp. For comparison of vBMD results between different software (aim 1) MECT 120 kVp and DECT synthetic 120 kVp data was used. For analyzing suitability of using different DECT energies for vBMD assessment (aim 2) all three DECT energies were used and results from each software was analyzed separately. Results: vBMD assessed with MW and IP, respectively correlated significantly for both the MECT (r=0.876; P&lt;0.001) and DECT (r=0.837; P&lt;0.001) scans, but the vBMD values were lower in using IP for vBMD assessment (8% and 14% lower for MECT and DECT, respectively; P=0.001). Regarding the different DECT energies, using MW for vBMD assessment showed significant correlations in vBMD results between 120 kVp and the two other energies (r=0.988 and r=0.939) and no significant differences in absolute vBMD values (P 0.05). The IP analysis as well showed significant correlation between 120 kVp and the other energies (r=0.769 and r=0.713, respectively), but differences in absolute vBMD values between the energies (P &lt;= 0.001). Conclusions: We show that the correlations between the vBMD derived from the two investigated software solutions were generally good but that absolute vBMD value did differ and might impact the clinical diagnosis of osteoporosis. Though small, our study data indicate that vBMD might be assessed in energies other than 120 kVp when using MW but not when using IP.

Background and purpose - A correct diagnosis is essential for the appropriate treatment of patients with atypical femoral fractures (AFFs). The diagnostic accuracy of radiographs with standard radiology reports is very poor. We derived a diagnostic algorithm that uses deep neural networks to enable clinicians to discriminate AFFs from normal femur fractures (NFFs) on conventional radiographs. Patients and methods - We entered 433 radiographs from 149 patients with complete AFF and 549 radiographs from 224 patients with NFF into a convolutional neural network (CNN) that acts as a core classifier in an automated pathway and a manual intervention pathway (manual improvement of image orientation). We tested several deep neural network structures (i.e., VGG19, InceptionV3, and ResNet) to identify the network with the highest diagnostic accuracy for distinguishing AFF from NFF. We applied a transfer learning technique and used 5-fold cross-validation and class activation mapping to evaluate the diagnostic accuracy.Results - In the automated pathway, ResNet50 had the highest diagnostic accuracy, with a mean of 91% (SD 1.3), as compared with 83% (SD 1.6) for VGG19, and 89% (SD 2.5) for InceptionV3. The corresponding accuracy levels for the intervention pathway were 94% (SD 2.0), 92% (2.7), and 93% (3.7), respectively. With regards to sensitivity and specificity, ResNet outperformed the other networks with a mean AUC (area under the curve) value of 0.94 (SD 0.01) and surpassed the accuracy of clinical diagnostics.Interpretation - Artificial intelligence systems show excellent diagnostic accuracies for the rare fracture type of AFF in an experimental setting.

Objective: To investigate the effect of the COVID-19 pandemic on denosumab persistence. Methods: All patients who initiated denosumab treatment at our outpatient care (osteoporosis/endocrinology unit) between 2016-2019 were included and date of injections were recorded from case records. Persistence was analyzed regarding 2-y persistence and 1-y persistence. Persistence was defined as a maximum interval of either 1) 8 months (m) [6+2m], or 2) 9m [6+3m]. Results: In total 171 patients were included. Mean age was 74.3+10.2 y (range 35-93 y) and 87% were women. Age and gender distribution did not differ significantly between the year of denosumab initiation. The 2-y persistence rate (8-m interval permitted) was lower in patients starting denosumab 2019 than those starting 2016-1018, i.e., 69% vs. 83%, p=0.044. No significant difference was seen analyzing 1-y persistence in the same groups (87 vs. 91%, p=0.410). When using a more liberal persistence definition, i.e., 9+m interval permitted, no difference was seen between 2 y persistence, 77 vs. 83%, p=0.341. Conclusion: The present study indicates that a higher number of patients got their injection later than recommended during the pandemic, but despite high pressure on our healthcare system and health concerns in the general population, no significant influence on denosumab persistence using a more liberal interval allowance was seen in this outpatient osteoporosis group.