Phantomless estimation of bone mineral density on computed tomography: a scoping review

Abstract

Background

Age-related conditions like osteoporosis have become more familiar with increasing global life expectancy. Osteoporosis is characterized by reduced bone mineral density and structural weakening of bone tissue that leads to a higher risk of fracture. Dual-energy X-ray absorptiometry is the current standard for diagnosing osteoporosis. Computed tomography provides an alternative for diagnosis, but traditional QCT involves the use of phantoms, which does not allow retrospective or opportunistic assessments of BMD. This study aims to provide an overview of the evidence and feasibility for emerging phantomless techniques for the estimation of BMD on CT.

Methods

A scoping review was conducted to evaluate the feasibility and effectiveness of phantomless BMD estimation using CT. A comprehensive search of Scopus and PubMed databases focused on literature published between 2010 and 2024. Search terms included combinations of “phantomless”, “BMD estimation”, and “CT”. Studies emphasizing accuracy, reliability, and clinical feasibility were included. The review identified 26 relevant studies examining methods of phantomless BMD estimation. The majority of the studies used internal anatomical references to calibrate BMD measurements. These methods demonstrated accuracy comparable to traditional phantom-based techniques. Limitations of the technique included variability in scanner types and inconsistencies caused by patient-specific factors like body composition and contrast agents.

Conclusion

Phantomless BMD estimation methods are a feasible approach to detecting osteoporosis. The possibility to be integrated into routine CT workflows make them an attractive option for opportunistic screening. Further research is necessary to refine methods, ensure consistent results across different clinical settings, and address outstanding issues such as scanner variability and the effects of contrast agents.

Key Points

• Phantomless CT BMD estimation methods are a feasible approach to detecting osteoporosis.

• Phantomless is an attractive option for opportunistic diagnosis and screening.

• Further studies need to address scanner variability and effects of contrast agents.

Citation Format

• Waqar A, Bazzocchi A, Aparisi Gómez MP. Phantomless estimation of bone mineral density on computed tomography: a scoping review. Rofo 2025; DOI 10.1055/a-2530-7790

Zusammenfassung

Hintergrund

Altersbedingte Erkrankungen wie Osteoporose sind mit der zunehmenden globalen Lebenserwartung häufiger geworden. Osteoporose ist durch eine verringerte Knochenmineraldichte und strukturelle Schwächung des Knochengewebes gekennzeichnet, was zu einem höheren Frakturrisiko führt. Die Dual-Energy-Röntgenabsorptiometrie ist der aktuelle Standard zur Diagnose von Osteoporose. Die Computertomografie bietet eine alternative Methode zur Diagnose, jedoch erfordert das traditionelle QCT die Verwendung von Phantomen, was retrospektive oder opportunistische Bewertungen der Knochenmineraldichte (BMD) nicht zulässt. Diese Studie zielt darauf ab, einen Überblick über die Evidenz und Machbarkeit aufkommender phantomloser Techniken zur Schätzung der BMD in der CT zu geben.

Methode

Eine Scoping-Übersicht wurde durchgeführt, um die Machbarkeit und Wirksamkeit der phantomlosen BMD-Schätzung mittels CT zu bewerten. Eine umfassende Suche in den Datenbanken Scopus und PubMed konzentrierte sich auf Literatur, die zwischen 2010 und 2024 veröffentlicht wurde. Die Suchbegriffe umfassten Kombinationen von „phantomlos“, „BMD-Schätzung“ und „CT“. Studien, die Genauigkeit, Zuverlässigkeit und klinische Machbarkeit betonten, wurden eingeschlossen. Die Übersicht identifizierte 26 relevante Studien, die Methoden der phantomlosen BMD-Schätzung untersuchten. Die Mehrheit der Studien verwendete interne anatomische Referenzen zur Kalibrierung der BMD-Messungen. Diese Methoden zeigten eine vergleichbare Genauigkeit zu traditionellen phantom-basierten Techniken. Zu den Einschränkungen der Technik gehörten die Variabilität der Scanner-Typen und die Inkonsistenzen, die durch patientenspezifische Faktoren wie Körperzusammensetzung und Kontrastmittel verursacht wurden.

Schlussfolgerung

Phantomlose BMD-Schätzungsmethoden sind ein machbarer Ansatz zur Erkennung von Osteoporose. Die Möglichkeit, in routinemäßige CT-Workflows integriert zu werden, macht sie zu einer attraktiven Option für opportunistisches Screening. Weitere Forschung ist notwendig, um Methoden zu verfeinern, konsistente Ergebnisse in verschiedenen klinischen Umgebungen sicherzustellen und verbleibende Probleme, wie die Scanner-Variabilität und die Auswirkungen von Kontrastmitteln, zu adressieren.

Kernaussagen

• Phantomlose CT-BMD-Schätzmethoden sind ein machbarer Ansatz zur Erkennung von Osteoporose.

• Phantomlos ist eine attraktive Option für opportunistische Diagnose und Screening.

• Weitere Studien müssen die Scanner-Variabilität und die Auswirkungen von Kontrastmitteln berücksichtigen.

Keywords

osteoporosis - phantomless - BMD - internal calibration - opportunistic

Abbreviations

BMD: Bone Mineral Density

CKD: Chronic Kidney Disease

CT: Computed Tomography

DXA: Dual Energy X-Ray Absorptiometry

DECT: Dual Energy Computed Tomography

FEA: Final Element Analysis

HU: Hounsfield Units

LDCT: Low Dose Computed Tomography

MDCT: Multidetector Computed Tomography

PB-QCT: Phantom-based Quantitative CT

PL-QCT: Phantomless Quantitative CT

PLvBMD: Phantomless Volumetric Bone Mineral Density

SAT: Subcutaneous Adipose Tissue

SDCT: Spectral Detector CT

Introduction

Due to improved living conditions and medical advancements, global aging has surged, leading to an increased prevalence of conditions like osteoporosis [1] [2]. Osteoporosis is defined as systemic reduced bone mineral density (BMD) and structural bone tissue deterioration, resulting in an increase in fracture risk [2] [3] [4].

Fragility fractures, which are common in osteoporosis, significantly increase mortality, disability, and healthcare costs [5] [6] [7]. In 2010, osteoporosis-related expenses within the European Union were approximately 37 billion euros, with over 70% of that amount being attributable to fracture-related costs [8].

While preventable and treatable, osteoporosis often remains undetected until advanced stages. Dual-energy X-ray absorptiometry (DXA) is the standard for diagnosing osteoporosis, but it has limitations [1] [3] [9]. DXA measures bone density using low-dose X-rays, providing a T-score that compares a patient’s BMD to that of a healthy young population.

However, variations in body composition, interference from soft tissue, bowel content, degenerative spine disease and vascular calcification can alter BMD measurements [2] [7] [10]. DXA measurements are areal (aBMD) and evaluate bone structure two-dimensionally [2] [6] [7] [11]. Osteoporosis predominantly affects trabecular bone over cortical bone, as it is more metabolically active [6] [7]. By overlooking the three-dimensional (3D) structure of bone, values generated from DXA may only partially reflect bone strength and varying fracture risks of individuals with similar values [1] [2] [7]. Quantitative CT (QCT) differentiates between cortical and trabecular bone and provides detailed 3D distribution of BMD [4]. Studies have shown it exhibits superior sensitivity in osteoporosis detection by eliminating confounding from osteophytes and vascular calcification [1]. Traditionally, the accuracy of results relies on calibrating volumetric BMD (vBMD) measurements with a reference phantom during the scan [1] [2] [7] (phantom-based QCT – PB-QCT). Calibration mitigates beam hardening and scattering, attributed to patient-specific factors such as body habitus [2].

However, PB-QCT entails logistical and financial challenges and does not allow retrospective or opportunistic assessments of BMD. External phantoms have some disadvantages, such as air gap artifacts, dependency on patient size, and other patient-moderated artifacts, resulting in repositioning errors affecting precision. When phantoms are scanned asynchronously, high stability of the scanner is required but may not always be present [2].

Phantomless QCT (PL-QCT) provides the opportunity to calculate BMD from any CT scan conducted for alternative purposes. Methods such as patient-specific internal calibration for estimating BMD have been proposed [1] [2] [7]. These rely on the HU of specific internal tissues as calibration references.

However, the extent of evidence on this topic and its feasibility need to be reviewed and clarified. A scoping review was conducted to reflect concepts, sources, and types of evidence and identify the gaps in knowledge for further research.

Methods

A scoping review was conducted per the PRISMA-ScR statement [12].

The search was executed using a three-step strategy. Initially, a preliminary literature search in PubMed and Scopus was performed using a block search to structure the final search parameters. Four categories formed the blocks: ‘Phantomless’, ‘Estimation’, ‘BMD’, and ‘CT’. The search period ranged from 2010 to 2024, and the language was set to English ([Table 1]).

Table 1 Block search carried out in Scopus and PubMed, July 2024.

	Scopus	PubMed
Phantomless	phantomless OR non-phantom OR opportunistic OR internal AND calibration	phantomless OR non-phantom OR opportunistic OR internal AND calibration
Estimation	estimation OR measurement OR assessment OR calculation OR calibration	estimation OR measurement OR assessment OR calculation OR calibration
Bone mineral density	bone AND mineral AND density OR BMD	bone AND mineral AND density OR BMD
Computed tomography	{computed tomography}	{computed tomography}

Titles and abstracts of the retrieved articles were screened based on the inclusion and exclusion criteria ([Table 2]). Subsequently, full-text articles were retrieved and assessed for eligibility. Data extracted from the selected studies included study design, Oxford level of evidence, aim/purpose, sample size, scanner type/s, mean age, outcome (feasibility/accuracy), limitations, future perspectives and additional information if relevant. This data was then systematically analyzed to identify common themes, discrepancies, and patterns in the research.

Table 2 Inclusion and exclusion criteria.

	Criteria
Inclusion	Studies published in English
	Randomized controlled trials, cohort studies, case-control studies, cross-sectional studies, validation studies, technical reports, and systematic reviews
	Accuracy, reliability, feasibility of phantomless BMD estimation; comparison with traditional methods
Exclusion	Studies involving animals or cadavers
	BMD estimation methods that require phantoms; BMD assessment using modalities other than CT
	Studies published before 2010.
	Letters to editor, case reports, expert opinions, literature reviews

Results

A total of 26 studies were included in this review ([Fig. 1]) after screening and the application of the inclusion and exclusion criteria. Detailed analysis is provided in [Table 3] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38].

Fig. 1 PRISMA flowchart.

Table 3 Detailed analysis of the reviewed articles.

Authors	Study Design	Oxford Evidence Level	Aim/Purpose	Sample Size	Scanner Type	Age	Contrast	Methodology Details	Outcome (Feasibility/Accuracy)	Limitations	Future Interpretation	Additional Information
Abdullayev et al (2018) [13]	Retrospective observational study	Level 3	To assess the effects of contrast enhancement on phantomless volumetric bone mineral density (vBMD)	56 datasets	Siemens Somatom Definition Flash, Philips Brilliance 64, Philips iCT 256, Philips Brilliance 6	The mean age is 64 years, age range is 20–80 years	Yes	L1 to L3. Calibration: paravertebral muscles (erector spinae muscle) and the subcutaneous fat tissue	Higher PLvBMD in contrast-enhanced scans by 8.6% compared to unenhanced scans; good interobserver variance and precision	Limited to abdominal CT; potential bias in heterogenous HU distribution in paraspinal muscles and subcutaneous fat tissue	PLvBMD can be improved with adjustment formulas for contrast-enhanced scans for osteoporosis screening	PLvBMD uses internal tissues (muscles and fat) as calibration points, improving practicality for opportunistic osteoporosis screening
Alacreu et al (2017) [14]	Retrospective cohort study	Level 3	Feasibility of early osteoporosis detection using abdominal CT scans compared with DXA for oncologic patients	326 datasets	Siemens Somatom Sensation 40-slice	Mean age 62 years, age range 50–74 years	Mixed 300 yes 26 no	Abdominal scans. L1 to L4. Calibration: ROIs were placed at axial cross-section of L1 from paraspinal muscles and adjacent subcutaneous fat	Highly sensitive thresholds: 160 HU for L1 for detecting osteoporosis with over 90% sensitivity; abdominal CT shows a high correlation with DXA	Limited external validation and geographic variation in osteoporosis prevalence; CT scans are primarily used for oncology patients	Further external validation of opportunistic screening methods, especially geographic-specific sensitivity thresholds	Abdominal CT scans can detect osteoporosis without additional radiation exposure or cost, improving early detection in high-risk patients
Bartenschlager et al (2022) [15]	Retrospective comparative study	Level 3	To assess accuracy errors in PL-QCT of lumbar spine BMD from routine CT scans	121 datasets	GE, Philips, Toshiba (various models not specified)	The mean age is 71; the age range is 65–85	No	Lumbar spine scans (L1 and L2). Calibration: Combinations air, blood, SAT, and skeletal muscle	BMD accuracy errors below 5% when using air and blood as calibration materials; highest errors with skeletal muscle as reference	Higher BMD errors when using skeletal muscle due to variability in fatty infiltration, especially in elderly patients	Air and blood are recommended for calibration in younger subjects; SAT and air are preferable in elderly subjects due to aortic calcifications	PL-QCT introduces some errors compared to PB-QCT but offers a practical approach for opportunistic osteoporosis screening
Bartenschlager et al (2023) [16]	Retrospective comparative	Level 3	Comparison of four calibration methods for BMD assessment in opportunistic screening	464 datasets	Standardized CT protocol (details not provided); various clinical CT scanners from routine settings	–	No	Comparison of methods: Three precalibrated (phantomless – CT values of air and of different human body tissues, blood, subcutaneous adipose tissue, skeletal muscle or cortical bone, were used in different combinations) and one based on internal material decomposition – voxel-specific calibration	PL-QCT performs well if pre-calibrated with the reference dataset; ΔBMD ranges from 0.1–2.7 mg/cm³ in the standardized dataset	Calibration without a reference dataset leads to higher error margins; some methods require prior dataset calibration	Further evaluation of phantomless calibration methods in routine CT screening for osteoporosis	PL-QCT provides a flexible solution for retrospective BMD analysis in clinical CT scans, but accuracy depends on the reference dataset and scanner stability
Boomsma et al (2016) [17]	Prospective cohort study	Level 2	To assess the use of internal references for CT density measurements in patients with metal-on-metal implants	50 participants	Philips Brilliance 40-slice, Philips Brilliance 64-slice, Philips Brilliance 128-slice	The mean age is 62, and the age range is 40–72	No	Metal hip replacements Calibration: SAT, skeletal muscle (psoas)	Excellent agreement between internal and phantom references for fat and muscle (ICC: 0.90 and 0.84)	Limited to non-contrast CT scans, metal implants create artifacts but good results in unaffected pelvis regions	Internal references are sufficient for CT density measurements without external phantoms, supporting opportunistic applications	Internal references for fat and muscle can replace external phantoms, allowing more flexibility in assessing bone and soft tissue density from CT scans without needing additional equipment
Budoff et al (2013) [18]	Retrospective observational study	Level 3	To evaluate the accuracy of PL-QCT BMD measurements on coronary artery calcium CT scans	4126 participants	GE LightSpeed VCT 64, various models from Siemens, Philips, Toshiba, and GE	Mean age 64 years	Mixed	Coronary artery calcium CT scans; correlation with phantom scans	PL-QCT BMD correlated highly with phantom-based measurements (r = 0.987); bias was 3.9% ± 1.4	Some limitations are due to the variations in calibration factors across different scanner models	PL-QCT BMD measurements may reduce the need for additional scans and radiation exposure when evaluating BMD during coronary artery calcium CT	PL-QCT BMD is a promising approach for reducing radiation exposure while maintaining accuracy in coronary CT scans
Eggermont et al (2019) [19]	Prospective cohort study	Level 2	To develop a PL-QCT method for fracture risk prediction in cancer patients with bone metastases	57 participants	Philips Brilliance Big Bore (Philips-1 and Philips-2), GE Optima CT580, Toshiba Aquilion/LB	Mean age 67 years	No	Advanced cancer patients, with bone metastases. Comparison: “air-fat-muscle” and “non-patient-specific” calibration (convert HU to BMD by averaging all calibration functions of all 26 scanned patients)	No significant differences exist between PB-QCT and PL-QCT air-fat-muscle calibration for FEA modelling accuracy	Limited to cancer patients with femoral bone metastases, and external validation is needed for general fracture risk prediction	PL-QCT can simplify fracture risk prediction using clinical CT scans without a calibration phantom	PL-QCT shows promise for increasing the practicality and affordability of finite element modelling in clinical settings for fracture risk prediction, especially in patients with bone metastases
Gruenewald et al (2022) [20]	Retrospective cohort study	Level 3	To assess the accuracy of (phantomless) DECT-derived volumetric BMD for predicting osteoporosis-associated fractures	92 participants	Siemens SOMATOM Force (third-generation dual-source DECT)	Age range: 19–103 years	No	L1 DECT	DECT-derived BMD of 93.7 mg/cm³ yielded 85.45% sensitivity and 89.19% specificity for fracture risk prediction	Limited by the retrospective design and reliance on a single DECT scanner model, external validation across diverse populations is needed	DECT-based volumetric BMD can be used to predict the 2-year fracture risk, and further research could focus on expanding this method to other populations.	DECT-based BMD assessments can replace DXA and QCT for assessing osteoporosis-related fracture risk, especially for opportunistic screening using existing CT scans
Gruenewald et al (2024) [21]	Retrospective comparative study	Level 3	To evaluate the use of DECT-derived BMD and other CT-based metrics for predicting distal radius fractures	263 participants	Siemens SOMATOM Force (third-generation dual-source DECT)	The mean age is 52 years.	No	Distal radius DECT	DECT-derived BMD was the strongest predictor for distal radius fractures with an AUC of 0.91 compared to cortical HU and trabecular HU	Limited access to DECT scanners restricts widespread clinical implementation and limited focus on cortical and trabecular HU and cortical thickness’s impact on prediction	DECT-based BMD could be used as an effective screening tool for osteoporosis and fracture risk stratification without calibration phantoms	DECT-derived BMD is a superior tool for opportunistic screening of bone health and fracture risk, with the potential for broader application in clinical practice
Guo et al (2024) [22]	Retrospective validation study	Level 3	To validate a semi-automatic PL-QCT system for measuring trabecular bone density in the proximal humerus	188 datasets	Neusoft GB18030 (NeuViz Prime 1.0)	The mean age is 53, and the age range is 18–65	No	Proximal humerus scans Calibration: Subcutaneous fat of the shoulder joint and trapezius muscle MEASUREMENT OF ROI/SEGMENTATION/CALCULATION BMD: Manual ROI and segmentation. Humeral greater tuberosity was selected on the axial plane as the center of the entire 3D ROI annotation. The axial ROI range was adjusted to be the largest possible range excluding cortical bone. Measurements on BMD based on these ROIS were semiautomatic	PL-QCT had good correlation with PB-QCT (R² = 0.97); bias of 1.0 mg/cm³	The study focused on the proximal humerus, and external validation in other anatomical regions is needed	The semi-automatic PL-QCT system shows promise for future applications in shoulder BMD measurement and fracture risk prediction	PL-QCT based on chest CT could enable more opportunistic screening for shoulder fractures using pre-existing scans, improving the diagnosis of osteoporosis
Habashy et al (2011) [23]	Retrospective comparative study	Level 3	To assess the feasibility of estimating BMD from diagnostic CT scans in children without using a calibration phantom	43 participants	General Electric Lightspeed Ultra 8 Detector CT scanner, Electric VCT XT 64 Detector unit	The median age is 13 years	Mixed	L1 and L2. Investigation of Bone Investigational Toolkit	PL-QCT measurements had a higher absolute standardized difference from PB-QCT by 14.3%	The use of diagnostic CT introduces measurement bias due to varying scan protocols and lack of external calibration	Further refinement in bias compensation for PL-QCT BMD could make routine diagnostic CT a reliable tool for bone assessment	PL-QCT methods could significantly enhance pediatric patient care by reducing additional radiation exposure while assessing BMD using diagnostic CT scans
Kaesmacher et al (2017) [24]	Retrospective comparative study	Level 3	To assess the effect of contrast enhancement on PL-QCT BMD calibration using MDCT (abdominal study of liver or kidney pathology)	46 participants	Philips iCT, Siemens Sensation 64 (MDCT)	The mean age is 64, and the age range is 45–70	Yes nonenhanced (NE), arterial (AR) and portal-venous (PV) contrast phase	L1-L3. Comparison: synchronous and asynchronous phantom calibration as well as internal calibration (paraspinal muscle, SAT)	PL-QCT BMD estimates were biased by contrast medium application, requiring adjustment with correction equations	Contrast medium alters BMD values, complicating measurements in contrast-enhanced scans, regardless of calibration method	Developing specific correction models for contrast-enhanced scans may allow opportunistic BMD assessments using routine clinical imaging	Both PB-QCT and PL-QCT can be adapted for contrast-enhanced MDCT, with adjustments for contrast-related biases
Lee et al (2017) [25]	Retrospective observational study	Level 3	To evaluate the precision of PL-QCT BMD calibration using air and blood/adipose tissue as internal references (spine and hip)	40 (sample)	GE LightSpeed VCT, GE BrightSpeed, Philips Brilliance, Siemens Sensation, Toshiba Aquilion	The mean age is 67; the age range is 41–86	Mixed	Spine and hip scans. Comparison phantom and internal calibration, using external air and either abdominal aortic blood tissue for assessment of the spine or pelvic visceral adipose tissue from the ischioanal fossa for assessment of the hip	PL-QCT showed equivalent results to phantom-based methods, with precision errors ≤0.5%	Precision errors were minor, but the manual selection of tissues for calibration introduces potential measurement variability	Automatic ROI selection systems could further improve the precision and reproducibility of phantomless BMD calibration	PL-QCT holds promise for routine clinical use, mainly when external calibration phantoms are impractical or unavailable
Liu et al (2022) [26]	Retrospective validation study	Level 3	To validate a new automatic PL-QCT system for measuring spinal BMD and diagnosing osteoporosis	63 participants	Siemens SOMATOM Force	The mean age is 61, and the age range is 20–90	N/A	Abdominal scans. L1-L3. Comparison PB-QCT and PL-QCT. Calibration: SAT and skeletal muscle (paravertebral)	The new automatic PL-QCT system had high precision (CV = 0.89%) and showed good agreement with DXA and PB-QCT for diagnosing osteoporosis	The relatively small sample size limited the study, and additional testing is needed to ensure the generalizability of the findings	Automatic PL-QCT systems could significantly enhance the convenience and precision of opportunistic osteoporosis screening in routine care	Automatic systems for PL-QCT are promising for reducing operator dependence and improving BMD measurement precision, potentially facilitating broader clinical use of opportunistic screening techniques
Mao et al (2016) [27]	Retrospective cohort study	Level 3	To assess the accuracy and precision of PL-QCT BMD measurements from coronary artery calcium scans	513 participants.	GE LightSpeed VCT	The mean age is 58 years	Mixed	Coronary artery calcium CT scans. Correlation with phantom scans	PL-QCT BMD correlated with PB-QCT with a 3.8% bias and similar precision errors	Limited to the thoracic spine, external validation in lumbar regions and across diverse scanners is needed	PL-QCT BMD measurements using coronary CT could be widely applied for opportunistic osteoporosis screening	PL-QCT BMD measurements with coronary artery calcium scans offer a dual benefit: assessing cardiovascular risk and bone density simultaneously without additional radiation exposure
Matheson et al (2024) [28]	Retrospective cohort study	Level 3	To introduce an internal calibration method for opportunistic BMD assessment and explore the error bounds	138 participants	Revolution GSI (GE Healthcare), Mindways Model 3 CT Calibration Phantom	The mean age is 56 years	No	Abdominal scans (CT kidneys, ureter bladder) Calibration: air, adipose, blood, cortical bone, and skeletal muscle	BMD errors of 0.06% to 0.02% were found with air, skeletal muscle, and cortical bone used as internal calibration materials	The error bounds need validation across other anatomical regions and with larger datasets	Future studies should focus on validating the error bounds and expanding internal calibration methods to other clinical applications.	The introduction of error bounds in internal calibration can improve the accuracy of opportunistic BMD measurement from routine clinical CT scans
Oh et al (2024) [29]	Retrospective cohort study	Level 3	To develop a deep learning-based method for opportunistic osteoporosis screening using PL-QCT	367 participants	Discovery CT 750HD (GE), Brilliance (Philips), SOMATOM Definition Edge (Siemens)	The mean age is 59; the age range is 50–88	Yes	L1 to L4. Paravertebral muscle and SAT. MEASUREMENT OF ROI / SEGMENTATION / CALCULATION BMD: The system generated end-to-end BMD using convolutional neural network (deep learning-based subtasks, complete process fully automated)	The deep learning model achieved a Pearson’s r of 0.852 for DXA BMD prediction from CT and demonstrated the feasibility of automated screening	Using contrast-enhanced abdominal CT scans may introduce bias, and further testing in non-oncology populations is needed	Future research should focus on improving the integration of deep learning models into clinical workflows for routine BMD screening	Deep learning offers a promising solution for opportunistic osteoporosis screening using routine CT scans without additional scans
Pan et al (2023) [30]	Retrospective cohort study	Level 3	To evaluate a fully automated PL-QCT BMD measurement system for chest LDCT scans	1175 participants.	Siemens SOMATOM Definition AS+, GE Optima CT540	The mean age is 59; the age range is 32–88	No	Chest scans. Paravertebral muscle and SAT. MEASUREMENT OF ROI / SEGMENTATION / CALCULATION BMD: The system generated end-to-end BMD using convolutional neural network (deep learning-based subtasks, complete process fully automated)	High correlation with PB-QCT (r = 0.896 in validation cohort I, r = 0.956 in validation cohort II); AUCs for osteoporosis were 0.876–0.903	The accuracy of BMD estimates varied by scanner type, and the agreement with PB-QCT was insufficient for interchangeable use in all settings	Fully automated opportunistic screening systems could be further developed for routine chest LDCT to detect low BMD and osteoporosis	The study highlights the potential for reducing osteoporosis screening gaps by integrating opportunistic BMD measurement into routine lung cancer screening CT scans
Prado et al (2021) [31]	Retrospective comparative study	Level 3	To develop and validate PL-QCT methods for fracture risk prediction using QCT and FEA	111 datasets	Siemens Sensation 64	The mean age is 67 years	No	FEA models. L3. Equations were developed based on the HU measured from various soft tissues and regions (air, fat, psoas muscle, and subject perimeter), and using multiple linear regression analyses	PL-QCT results highly correlated with phantom-based measurements for vertebral fracture prediction	Limited to vertebral fractures, more studies are needed to validate across different fracture types and anatomical regions	PL-QCT could simplify clinical fracture risk prediction by removing the need for calibration phantoms during routine scans	PL-QCT in QCT/FEA could enhance clinical fracture risk prediction and facilitate its broader adoption in routine care without the logistical burden of phantoms.
Prado et al (2024) [32]	Retrospective observational study	Level 3	To validate vertebral fracture risk thresholds using PL-QCT/FEA modelling	108 datasets	Siemens Sensation-64	The mean age is 68 years	No	FEA models. L3. Equations were developed based on the HU measured from various soft tissues and regions (air, fat, psoas muscle, and subject perimeter)	Strong prediction correlations between PB-QCT and PL-QCT (R² > 0.95); accurate fracture risk thresholds	The study focused on vertebral fractures; validation in other fracture types is needed	PL-QCT /FEA modelling is a promising method for predicting vertebral strength and fracture risk in clinical settings	PL-QCT can be used prospectively and retrospectively to assess fracture risk, reducing the need for calibration phantoms
Szyszko et al (2024) [33]	Retrospective comparative study	Level 3	To compare phantom-based and phantomless calibration methods for hip fracture risk prediction	41 datasets	GE HiSpeed CT/I	The mean age is 67; the age range is 56–84	No	Spine and hip scans Calibration: air, adipose, and muscle tissues MEASUREMENT OF ROI / SEGMENTATION / CALCULATION BMD: Manual ROI selection and volume segmentation. Automatic calculation of FEA	Excellent correlation between PB-QCT and PL-QCT (R² > 0.99); RMSRE values of 1.99–4.02%	The manual selection of tissues introduces variability in PL-QCT, affecting reproducibility	PL-QCT shows promise for hip fracture prediction, but further refinement is needed for clinical adoption	PL-QCT can reduce logistical challenges and costs in fracture risk assessment, but automation could improve consistency
Therkildsen et al (2018) [34]	Comparative observational study	Level 2	To determine agreement between PL-QCT and PB-QCT BMD measurements in patients with CKD	149 participants	Siemens SOMATOM Definition Flash	The mean age is 54; the age range is 23–72	Yes	Calibration: posterior subcutaneous fat on the right side and the paraspinal muscle group on the left side	PB-QCT BMD was systematically higher than PL-QCT measurements, with a slight absolute difference (3.3 mg/cm³)	Intra- and interoperator variability was higher in PL-QCT, limiting its precision for clinical use	Further research is needed to improve the precision of PL-QCT and validate its clinical feasibility	PL-QCT is a promising tool for opportunistic screening, but its accuracy in patients with CKD requires further validation
Tong et al (2024) [35]	Prospective observational study	Level 2	To assess the feasibility of using DECT material decomposition for opportunistic osteoporosis screening	518 participants	GE Revolution CT (256-row)	The mean age is 63; the age range is 32–87	Yes	DECT	A high correlation between DECT-derived and QCT-derived BMD (AUC 0.963 for osteoporosis diagnosis)	BMD assessment may be limited in contrast-enhanced DECT scans, requiring further refinement to eliminate iodine interference	DECT material decomposition could offer a non-invasive, opportunistic screening tool for osteoporosis using existing contrast-enhanced scans	DECT-based BMD screening has the potential to significantly improve early detection of osteoporosis without requiring additional scans or radiation
Van Hedent et al (2019) [36]	Retrospective cohort study	Level 3	To evaluate the accuracy of spectral detector CT (SDCT) for opportunistic BMD assessment	20 participants + phantom	Philips IQon SDCT	The age range is 56–86 years	No	SDCT	SDCT showed high sensitivity (100%) and moderate specificity (73%) compared to DXA, with excellent correlation for K2HPO₄ measurements	Further validation in prospective clinical trials is needed to confirm SDCT’s feasibility for routine clinical BMD screening	SDCT could provide a more accurate method for opportunistic BMD screening than traditional QCT or DXA	SDCT combines the benefits of dual-energy CT and volumetric BMD analysis, improving the detection of low bone density without additional calibration phantoms
Weaver et al (2015) [37]	Retrospective validation study	Level 3	To validate PL-QCT lumbar spine vBMD measurement and assess correlation with age and fracture risk	50 participants.	GE LightSpeed VCT, 64-slice	The age range is 60–76 years	Mixed	Lumbar spine and abdominal scans. Calibration: ROIs were placed on the L1–L5 vertebrae, right psoas major muscle, and anterior subcutaneous fat. Regression model built	The strong correlation between PL-QCT and PB-QCT (R² = 0.87); validated method for assessing age and fracture incidence	Lack of external validation beyond motor vehicle collision injury cases and reliance on thoracic and lumbar spine regions for calibration	PL-QCT in trauma settings could facilitate broader research in bone health assessment, particularly in aging populations.	PL-QCT can be widely used in trauma research and clinical settings to assess lumbar bone health from routine CT scans without additional calibration
Winsor et al (2021) [38]	Retrospective cohort study	Level 3	To evaluate tissue-based PL-QCT calibration for finite element analysis of femoral fractures	258 datasets	GE LightSpeed VCT, Discovery CT750HD, Optima 580, Revolution GSI	The age range is 50–95 years	Yes	Calibration: adipose tissue, aortic blood, skeletal muscle, urine, and air	PL-QCT calibration produced femoral strength and BMD biases of less than 0.07%, demonstrating strong consistency with PB-QCT calibration	The variability of PL-QCT calibration across different tissue types (fat, muscle, blood) introduces some errors in extreme cases	Future studies should focus on refining the tissue-based phantomless calibration for broader use in clinical fracture risk prediction	PL-QCT calibration was cost-effective and versatile, offering reliable bone quality assessments across different patients

Twenty-two studies consisting of observational or retrospective cohort designs had Oxford Level 3 evidence. Additionally, there were four Level 2 studies, comprised of prospective cohorts.

The searches yielded studies encompassing a broad range of types of protocols: abdominal scans [14] [26] [28], chest [30], lumbar spine [15] [33] [37], coronary artery calcium CT scans [18] [27], contrast studies [13] [24], proximal humerus [22], spine and hip [25] [33]. The population demographics also included children in one of the studies [23]. The clinical settings were diverse: patients with implants [17], oncologic patients without bone metastases [14] and with bone metastases [19], patients with chronic kidney disease [34], and trauma patients [37]. The geographic locations of the articles were diverse. Additionally, several studies explored the effect of PL-QCT on final element analysis (FEA) outcomes for fracture risk [19] [31] [32] [38]. Three studies explored the application of automated methods [26] [29] [30].

Three studies focused on the use of dual-energy CT (DECT) as an inherently phantomless technique [20] [21] [35] and one on spectral detector CT (SDCT) [36].

The smallest series had 20 participants [36], and the largest had 4126 [15]. The studies utilized a diverse array of CT scanners.

The overall quality of research regarding BMD measurement using CT imaging is moderate. Studies are characterized by a focus on improving diagnostic accuracy, clinical applicability, and technological innovation. However, there are variations in study quality based on factors such as study design, sample size, and how well limitations are addressed.

Discussion

In recent years, the use of CT imaging to measure BMD has gained considerable interest for improving osteoporosis diagnosis and predicting fracture risk. DXA has long been considered the reference standard for assessing bone health. Advancements and innovations in CT technology have offered new possibilities [1] [6] [39]. Techniques that do not require the use of phantoms, such as internally calibrated QCT, DECT, and deep learning-based methods have shown great potential and significantly improved accessibility, accuracy, and automation. This is a significant benefit, as external phantoms can be cumbersome and inconvenient to use in routine clinical practice. Despite these promising developments, specific patterns and gaps remain in current research.

Recent research [15] [16] [28] [29] [30] [37] [38] has consistently shown that PL-QCT can produce comparable results to PB-QCT without the need for additional equipment.

Therkildse et al. [34] suggested that PL-QCT methods can closely align with PB-QCT methods, reporting discrepancies around 3.3 mg/cm³ [34]. However, authors found that intra- and interoperator variability were higher in PL-QCT, thus limiting its precision for clinical use.

Bartenschlager et al. [15] reported that the PL-QCT approach used for opportunistic screening creates additional BMD accuracy errors of 2% or more (compared to PB-QCT) depending on the used internal reference tissues, with the worst results being obtained with skeletal muscle. Using a CT value of 150 HU, a typical value of trabecular bone, simulated BMD accuracy errors for most calibration material combinations containing air as one of the two base materials were below 5% (<6 mg/cm³). The lowest errors were found for the combination of blood and air (<2 mg/cm³). The combination of blood and skeletal muscle resulted in higher errors (>10.5% or >12 mg/cm³) and was discouraged.

Authors suggested selecting calibration materials according to age or body composition, for example air and blood of the aorta in younger patients (with no aortic calcifications) and air and subcutaneous fat in elderly patients with vascular calcifications. Skeletal muscle is discouraged in the elderly population, because of the different degrees of fatty infiltration.

In a different study [16], authors compared four different calibration methods (three precalibrated and one based on internal material decomposition – voxel-specific calibration) with a mean CT calibration method (from the reference dataset – similar to an asynchronous calibration) and concluded that differences between PB-QCT and PL-QCT were small (<3 mg/cm³). They noted that the performance of the voxel-specific calibration compared to precalibrated methods was inferior compared to the other methods because one input parameter is the CT value of cortical bone, a very inaccurate measurement in the spine.

Matheson et al. [28] reported BMD errors of 0.06% to 0.02% when air, skeletal muscle, and cortical bone were used as calibration materials, in their study focused on the lumbar spine. However, their population sample is limited in size (n=138), so the error bounds need validation with larger samples and possibly other anatomical regions.

Additionally, Abdullayev et al. [13] reported that compared with the results of PB-QCT measurements reported in the literature, the vBMD changes associated with contrast enhancement obtained with PL-QCT were relatively moderate with an increase of 8.6% on average being considered an acceptable variation for clinical use. The phase of contrast did not appear to affect the results, but authors acknowledge that there is a potential bias in the heterogenous HU distribution in paraspinal muscles and subcutaneous fat tissue.

Several studies were conducted to validate objective processes for selecting tissues for use as the basis of PL-QCT to enable patient-specific final element analysis (FEA)-derived vertebral [25] [31] [32] and hip [25] [33] [38] fracture risk prediction.

Prado et al. [31] [32] reported strong prediction correlations between PB-QCT and PL-QCT using air, fat, the psoas muscle, and waist circumference (R² > 0.95), with accurate vertebral fracture risk thresholds. However, this is limited to vertebral fractures, and validation in other types of insufficiency fractures is needed.

Lee et al. [25] used air and either aortic blood or hip adipose tissue as internal calibration features and compared to PB-QCT with the conclusion that there were non-significant differences between measurements of aBMD at the hip (total hip and femoral neck), trabecular vBMD at the spine, and vertebral and femoral strength by FEA.

Winsor et al. [38] studied the use of different tissues for calibration. For each patient, they created FEA models (three were PB-QCT (synchronous or asynchronous) and two were PL-QCT (air, aortic blood, skeletal muscle and air and adipose)) and concluded that the air, aortic blood, skeletal muscle combination was the best for PL-QCT, with regard to BMD (resulted in an increase of only 4% for PL-QCT compared to PB-QCT) and femoral strength analysis through FEA.

Szyszko et al. [33] used air, muscle, and adipose tissues. Comparison of PB-QCT and PL-QCT demonstrated that comparison of FEA models obtained with the two methods showed excellent correlations (R² >99) for hip fracture risk prediction.

Fat and muscle are widely regarded as the most reliable internal calibration materials [13] [17] [33]. Using these tissues as calibration standards results in minimal errors, making them practical for routine clinical use. Air and soft tissue, such as lung parenchyma or blood, offer alternative calibration points, especially in regions like the thorax and abdomen. These tissues cover a broad range of densities, providing accurate calibration across varying bone densities. However, variability in patient body composition and the use of contrast agents can affect the reliability of these methods, highlighting the need for further refinement and standardization [13].

Another significant advancement in BMD measurement, without the need of a phantom, has been the application of DECT. Several studies [20] [21] [35] [40] underscore the accuracy of DECT over traditional single-energy CT methods.

As an important difference from conventional CT, DECT utilizes two energy levels. Single-energy CT (conventional CT) assesses the sum of the photoelectric effect and Compton scatter as global attenuation. Factors such as tube attenuation voltage levels per se, scanner settings, and protocols have an important effect on BMD measurements. Dual-energy CT (DECT) can separately assess these two interactions, reducing the impact of these effects. By taking advantage of the differential X-ray absorption characteristics of specific substances, material decomposition can be achieved, and other tissues can be effectively differentiated from bone. Different types of materials are used to represent the attenuation of specific tissues, enabling quantitative analysis of substances.

The main components of bone include large amounts of bone minerals such as Calcium (Ca) and Hydroxyapatite (HA), as well as other components such as red bone marrow, yellow bone marrow (mainly fat), collagen, and water. In traditional CT scans, fat marrow density affects HU values within individual voxels, as each voxel captures a mixture of bone and marrow fat [20].

Hydroxyapatite as a calcium and phosphate compound is thought to better reflect BMD information, but it is still unproven if the relative content of Ca obtained through the application of material separation techniques is capable of effectively reflecting bone density information [35].

The analysis by Grueneweld et al. [20] of non-enhanced DECT lumbar spine studies yielded an optimal DECT-based BMD threshold of 93.7 mg/cm³ to distinguish patients who sustained a fracture during a 2-year follow-up period from patients who did not, in opposition to the established QCT value of 80 mg/cm³ (according to ACR). The higher threshold they obtained can in part be attributed to the effect of the fat error, in line with ex vivo studies that reported that DECT-derived BMD values differ significantly from those obtained by QCT (p < 0.001) and are found to be closer to true HA concentrations [41].

Authors were able to minimize accuracy errors for a more accurate assessment of true trabecular BMD, ultimately providing sensitivity and specificity rates of 85.45% and 89.19%, respectively, which are higher than previously reported, making DECT particularly promising for clinical use.

Moreover, DECT’s ability to predict fracture risk across various bone regions—including the thoracic spine [20] and the distal radius [21]—suggests broad applicability. In the distal radius fracture prediction, DECT outperformed conventional CT metrics, such as cortical thickness and HU values [21].

The DECT material separation technique offers multiple base materials, including iodine. Through the selection of any two base materials and completion of base-material imaging, effective material separation and relative quantification can be achieved. This allows for the opportunistic assessment of BMD in the context of use of contrast.

Tong et al. [35] explored the feasibility of opportunistic screening for osteoporosis using enhanced CT based on the material decomposition technique in fast-switching DECT, compared to conventional QCT. The authors paired HA and iodine and Ca and iodine and found that there were no statistically significant differences in HA-iodine and Ca-iodine density values of L1–L3 vertebral bodies on triphasic enhanced CT, and that strong positive correlations were found between HA-iodine and BMD (r=0.9472), and between Ca-iodine and BMD (r=0.9470). They found good agreement between the calculated BMD-DECT and the reference BMD-QCT, concluding that HA (on HA-iodine images) and Ca (on Ca-iodine images) density measurements generated using the material decomposition technique in DECT have good diagnostic performance in assessing BMD and diagnosing osteoporosis.

In addition to diagnosing osteoporosis and predicting fracture risk, DECT can provide additional diagnostic information, such as coronary atherosclerosis detection and body composition quantification.

A study using SDCT, as a different means to measure BMD, showed high sensitivity (100%) and moderate specificity (73%) compared to DXA [36]. However, further validation in prospective clinical trials is needed to confirm SDCT’s feasibility for routine clinical BMD screening.

Deep learning and AI-based systems represent a significant advancement in BMD measurement techniques, offering the potential for automation and enhanced efficiency. Several studies [22] [29] [33] highlight the strong performance of AI-based models in automatically measuring BMD from routine CT scans. These deep learning systems have been shown to accurately diagnose osteoporosis by analyzing vertebrae and other bones without external calibration devices. For example, Guo et al. [22] demonstrated that AI-based systems achieved high levels of accuracy with minimal human intervention, thereby improving the efficiency of osteoporosis screening and reducing operator-related variability, However, their study was focused on the proximal humerus, and validation in other anatomical locations is needed. In fact, Szyszko et al. [33] noted that manual selection of tissues introduces variability in PL-QCT, which affects reproducibility.

AI-based systems reduce the time required for image analysis, making them particularly useful for large-scale screening programs. As AI systems evolve, their integration into routine BMD assessments could significantly reduce the burden on healthcare professionals while enhancing diagnostic precision.

Despite the evolution of phantomless calibration methods, traditional PB-QCT remains essential in specific high-risk patient populations. PB-QCT is crucial in patients with chronic conditions, such as chronic kidney disease or cancer, as even slight variations in BMD measurements, can significantly impact fracture risk assessment. Therkildsen et al. [34] emphasized the importance of accurate BMD assessments in patients with chronic kidney disease. In CKD, bone metabolism is disrupted, leading to significant cortical bone loss. Similarly, Alacreu et al. [14] stressed the need for precise BMD readings in cancer patients, where bone health is often compromised due to metastasis. Eggermont et al. [19] further emphasized that in populations with delicate bone health, accurate readings ensure that slight deviations in BMD do not lead to misjudgments in clinical decision-making, thus preventing under- or overestimation of fracture risk.

PB-QCT remains a valuable tool in cases where great accuracy is necessary – for example, in cases where precise monitoring of treatment effects is paramount [19].

The overall quality of research in BMD measurement using CT imaging is moderate, with multiple studies being retrospective/observational. However, several studies incorporate cross-validation techniques and longitudinal designs. For example, Prado et al. [31] used an additional 31 PB-QCT scans to validate the fracture risk using PL-QCT and FEA. Gruenewald et al. [20] included a 2-year follow-up period to assess the effectiveness of DECT in predicting osteoporosis-related fracture risk over time. The studies by Guo et al. [22] and Liu et al. [26] showcase how AI and machine learning techniques can automate BMD measurement, reduce human error, and improve diagnostic accuracy. These studies follow best practices in AI research, such as using large, well-labelled datasets and conducting rigorous testing to validate the performance of the models.

A standard limitation is the restricted generalizability of findings due to specific patient populations or bone regions. For example, research by Bartenschlager et al. [15] and Weaver et al. [37] focuses primarily on the lumbar spine, which limits the applicability of their results to other bone regions, such as the hip or femur, where BMD measurement is equally important. Similarly, some studies focus on patients with CKD or cancer whose bone density characteristics may differ from the general population [14] [19] [34]. Additionally, some studies focus on relatively small and homogeneous populations [19] [33]. This limits how these methods would perform across different demographic groups with the need for broader validation across diverse clinical settings and patient demographics [26].

Some studies are limited by small sample sizes, thus reducing the statistical power of their conclusions. For instance, the studies by Lee et al. [25] and Habashy et al. [23] include fewer than 100 participants, making it difficult to generalize their findings to larger populations. Small sample sizes also increase the risk of skewed results due to outliers.

Several studies point out that the manual selection of tissues for calibration in PL-QCT introduces potential measurement variability, thereby introducing some errors [25] [33] [38].

One recurring issue in the research is the variability in scanner types used across different studies. Variability in scanner types and calibration methods arises from differences in hardware, software, and imaging protocols [13] [16] [33]. Other studies [15] [19] emphasized that tube voltage and scanner model variations can lead to inconsistent BMD measurements, highlighting the need for standardized calibration protocols. Some studies address this variability by testing across multiple scanner types [15] [18] [22], with a similar conclusion that standardization across different scanner types is needed [18]. This may limit the reliability of conclusions.

Future directions

Although phantomless calibration methods offer practical advantages, they can introduce variability in BMD measurements, particularly in trabecular bone [15] [33]. Future research should focus on refining these methods by developing more sophisticated algorithms that account for individual variations in tissue composition. Moreover, ensuring consistency across different scanner types is essential for reducing variability in BMD measurements.

Another gap in the research is the need for comprehensive exploration of confounding variables, such as the impact of contrast agents on BMD measurements. Standardized correction equations to adjust for the effects of contrast agents on BMD measurements is needed [13] [23] [24]. While some studies have addressed this issue, these corrections have yet to be universally validated across different scanner types and clinical settings. Future research should focus on developing standardized protocols for adjusting BMD measurements on contrast-enhanced CT scans, ensuring consistency in clinical practice. DECT offers a promising alternative in this respect [35].

Most studies focus on the lumbar and thoracic spine, but there is a need for further validation of these methods across other bone regions, such as the hip, femur, and shoulder. Since different bones have varying densities and structures, techniques that work well in one area may be less effective in others. Moreover, studies should include more diverse patient populations, such as individuals with diabetes, obesity, or other comorbidities, to ensure these methods are effective across various clinical scenarios.

Another area for improvement is the need for long-term follow-up. This is particularly evident in studies focusing on AI, where initial results are promising, but the long-term reliability of these methods still needs to be determined.

Integrating imaging data with clinical decision tools, such as the FRAX score, has shown promise with respect to improving the accuracy of fracture risk predictions. Future research should develop hybrid models combining CT-based BMD data with clinical risk assessment tools to provide more comprehensive and individualized risk assessment. This integration could lead to earlier diagnosis and more personalized treatment strategies, thus improving patient outcomes.

Conclusion

Phantomless CT BMD measurement techniques have reasonable accuracy and clinical applicability. However, challenges remain, including addressing precision errors and standardizing methods. Future studies must address limitations such as small sample sizes and variability across scanner types to ensure reliable clinical application. DECT offers promising results, and integrating AI presents exciting possibilities. Addressing these gaps will make phantomless CT BMD-derived measurement techniques more reliable for diagnosing osteoporosis and predicting fracture risk.

Conflict of Interest

The authors declare that they have no conflict of interest.

• References

• 1 Mallio CA, Vertulli D, Bernetti C. et al. Phantomless Computed Tomography-Based Quantitative Bone Mineral Density Assessment: A Literature Review. Applied Sciences 2024; 14: 1447
• 2 Lee YH, Kim JJ, Jang IG. Patient-Specific Phantomless Estimation of Bone Mineral Density and Its Effects on Finite Element Analysis Results: A Feasibility Study. Comput Math Methods Med 2019; 2019: 4102410
• 3 Aparisi Gómez MP. Nonspinal Fragility Fractures. Semin Musculoskelet Radiol 2016; 20: 330-344
• 4 Guerri S, Mercatelli D, Aparisi Gómez MP. et al. Quantitative imaging techniques for the assessment of osteoporosis and sarcopenia. Quant Imaging Med Surg 2018; 8: 60-85
• 5 Sözen T, Özışık L, Başaran NÇ. An overview and management of osteoporosis. Eur J Rheumatol 2017; 4: 46-56
• 6 Bazzocchi A, Isaac A, Dalili D. et al. Imaging of Metabolic Bone Diseases: The Spine View, Part I. Semin Musculoskelet Radiol 2022; 26: 478-490
• 7 Booz C, Noeske J, Albrecht MH. et al. Diagnostic accuracy of quantitative dual-energy CT-based bone mineral density assessment in comparison to Hounsfield unit measurements using dual x-ray absorptiometry as standard of reference. European Journal of Radiology 2020; 132: 109321
• 8 Rashki Kemmak A, Rezapour A, Jahangiri R. et al. Economic burden of osteoporosis in the world: A systematic review. Med J Islam Repub Iran 2020; 34: 154
• 9 Bazzocchi A, Ferrari F, Diano D. et al. Incidental Findings with Dual-Energy X-Ray Absorptiometry: Spectrum of Possible Diagnoses. Calcif Tissue Int 2012; 91: 149-156
• 10 Bazzocchi A, Diano D, Ponti F. et al. A 360-degree overview of body composition in healthy people: relationships among anthropometry, ultrasonography, and dual-energy x-ray absorptiometry. Nutrition 2014; 30: 696-701
• 11 Bazzocchi A, Ponti F, Albisinni U. et al. DXA: Technical aspects and application. European Journal of Radiology 2016; 85: 1481-1492
• 12 Tricco AC, Lillie E, Zarin W. et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med 2018; 169: 467-473
• 13 Abdullayev N, Neuhaus V-F, Bratke G. et al. Effects of Contrast Enhancement on In-Body Calibrated Phantomless Bone Mineral Density Measurements in Computed Tomography. J Clin Densitometry 2018; 21: 360-366
• 14 Alacreu E, Moratal D, Arana E. Opportunistic screening for osteoporosis by routine CT in Southern Europe. Osteoporos Int 2017; 28: 983-990
• 15 Bartenschlager S, Dankerl P, Chaudry O. et al. BMD accuracy errors specific to phantomless calibration of CT scans of the lumbar spine. Bone 2022; 157: 116304
• 16 Bartenschlager S, Cavallaro A, Pogarell T. et al. Opportunistic Screening With CT: Comparison of Phantomless BMD Calibration Methods. J Bone Miner Res 2023; 38: 1689-1699
• 17 Boomsma MF, Slouwerhof I, van Dalen JA. et al. Use of internal references for assessing CT density measurements of the pelvis as replacement for use of an external phantom. Skelet Radiol 2015; 44: 1597-1602
• 18 Budoff MJ, Malpeso JM, Zeb I. et al. Measurement of phantomless thoracic bone mineral density on coronary artery calcium CT scans acquired with various CT scanner models. Radiology 2013; 267: 830-836
• 19 Eggermont F, Verdonschot N, van der Linden Y. et al. Calibration with or without phantom for fracture risk prediction in cancer patients with femoral bone metastases using CT-based finite element models. PLoS ONE 2019; 14
• 20 Gruenewald LD, Koch V, Martin SS. et al. Diagnostic accuracy of quantitative dual-energy CT-based volumetric bone mineral density assessment for the prediction of osteoporosis-associated fractures. European Radiology 2022; 32: 3076-3084
• 21 Gruenewald LD, Booz C, Gotta J. et al. Incident fractures of the distal radius: Dual-energy CT-derived metrics for opportunistic risk stratification. European Journal of Radiology 2024; 171
• 22 Guo DM, Weng YZ, Yu ZH. et al. Semi-automatic proximal humeral trabecular bone density assessment tool: technique application and clinical validation. Osteoporos Int 2024; 35: 1049-1059
• 23 Habashy AH, Yan X, Brown JK. et al. Estimation of bone mineral density in children from diagnostic CT images: a comparison of methods with and without an internal calibration standard. Bone 2011; 48: 1087-1094
• 24 Kaesmacher J, Liebl H, Baum T. et al. Bone Mineral Density Estimations From Routine Multidetector Computed Tomography: A Comparative Study of Contrast and Calibration Effects. J Comput Assist Tomogr 2017; 41: 217-223
• 25 Lee DC, Hoffmann PF, Kopperdahl DL. et al. Phantomless calibration of CT scans for measurement of BMD and bone strength—Inter-operator reanalysis precision. Bone 2017; 103: 325-333
• 26 Liu L, Si M, Ma H. et al. A hierarchical opportunistic screening model for osteoporosis using machine learning applied to clinical data and CT images. BMC Bioinformatics 2022; 23: 63
• 27 Mao SS, Li D, Luo Y. et al. Application of quantitative computed tomography for assessment of trabecular bone mineral density, microarchitecture and mechanical property. Clin Imaging 2016; 40: 330-338
• 28 Matheson BE, Neeteson NJ, Boyd SK. Establishing error bounds for internal calibration of quantitative computed tomography. Medical Engineering & Physics 2024; 124: 104109
• 29 Oh J, Kim B, Oh G. et al. End-to-End Semi-Supervised Opportunistic Osteoporosis Screening Using Computed Tomography. Endocrinol Metab (Seoul) 2024; 39: 500-510
• 30 Pan Y, Zhao F, Cheng G. et al. Automated vertebral bone mineral density measurement with phantomless internal calibration in chest LDCT scans using deep learning. Br J Radiol 2023; 96: 20230047
• 31 Prado M, Khosla S, Chaput C. et al. Opportunistic application of phantom-less calibration methods for fracture risk prediction using QCT/FEA. Eur Radiol 2021; 31: 9428-9435
• 32 Prado M, Khosla S, Giambini H. Vertebral Fracture Risk Thresholds from Phantom-Less Quantitative Computed Tomography-Based Finite Element Modeling Correlate to Phantom-Based Outcomes. J Clin Densitom 2024; 27: 101465
• 33 Szyszko JA, Aldieri A, La Mattina AA. et al. Phantomless calibration of CT scans for hip fracture risk prediction in silico: Comparison with phantom-based calibration. PLoS ONE 2024; 19
• 34 Therkildsen J, Thygesen J, Winther S. et al. Vertebral Bone Mineral Density Measured by Quantitative Computed Tomography With and Without a Calibration Phantom: A Comparison Between 2 Different Software Solutions. Journal of Clinical Densitometry 2018; 21: 367-374
• 35 Tong X, Fang X, Wang S. et al. Opportunistic screening for osteoporosis using enhanced images based on dual-energy computed tomography material decomposition: a comparison with quantitative computed tomography. Quantitative Imaging in Medicine and Surgery 2024; 14: 352-364
• 36 Van Hedent S, Su KH, Jordan DW. et al. Improving Bone Mineral Density Assessment Using Spectral Detector CT. J Clin Densitom 2019; 22: 374-381
• 37 Weaver AA, Beavers KM, Hightower RC. et al. Lumbar Bone Mineral Density Phantomless Computed Tomography Measurements and Correlation with Age and Fracture Incidence. Traffic Injury Prevention 2015; 16: S153-S160
• 38 Winsor C, Li X, Qasim M. et al. Evaluation of patient tissue selection methods for deriving equivalent density calibration for femoral bone quantitative CT analyses. Bone 2021; 143: 115759
• 39 Aparisi Gómez MP, Isaac A, Dalili D. et al. Imaging of Metabolic Bone Diseases: The Spine View, Part II. Semin Musculoskelet Radiol 2022; 26: 491-500
• 40 Gruenewald LD, Koch V, Yel I. et al. Association of Phantomless Dual-Energy CT-based Volumetric Bone Mineral Density with the Prevalence of Acute Insufficiency Fractures of the Spine. Academic Radiology 2023; 30: 2110-2117
• 41 Koch V, Hokamp NG, Albrecht MH. et al. Accuracy and precision of volumetric bone mineral density assessment using dual-source dual-energy versus quantitative CT: a phantom study. Eur Radiol Exp 2021; 5: 43

Publication History

Received: 13 December 2024

Accepted after revision: 25 January 2025

Article published online:
05 March 2025