Multi-classification of Grading Stages for Osteoporosis Using Abdominal Computed Tomography with Clinical Variables: Application of Deep Learning with a Convolutional Neural Network

doi:10.21203/rs.3.rs-3440051/v1

Download PDF

Research Article

Multi-classification of Grading Stages for Osteoporosis Using Abdominal Computed Tomography with Clinical Variables: Application of Deep Learning with a Convolutional Neural Network

https://doi.org/10.21203/rs.3.rs-3440051/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: Osteoporosis is a significant global health concern and is often undetected until a fracture occurs. To improve early detection, a deep learning (DL) model was developed to classify osteoporosis stages using abdominal computed tomography (CT) scans.

Materials and Methods: This study retrospectively collected data from scanned contrast-enhanced abdominal CT. A total of 3,012 acquired CT scan data DL models were constructed for using image data, demographic information, and multi-modality data, respectively. The three groups were defined according to T-score [normal (T-Score ≥ –1.0), osteopenia (–2.5 < T-Score < –1.0), and osteoporosis (T-Score ≤ –2.5)] derived from dual-energy X-ray absorptiometry and assessed by a qualified radiologist. In the DL process, we used the Gradient-Weighted Class Activation Mapping (Grad-CAM) technique to identify features and accurately interpret clinical areas.

Results: Of the 3,012 data sets, the results of the multimodal dataset models showed the highest area under the receiver operating characteristic curve (AUC) (0.94) and accuracy (ACC) (0.80), while the image data model showed an AUC of 0.93 and an ACC of 0.79. The model using demographic information independently showed the worst performance with an AUC of 0.85 and an ACC of 0.68. The sensitivity and specificity for the multimodal model are 0.80 and 0.90, respectively, while the demographic data model scored 0.69 and 0.84. In addition, the Grad-CAM identified informative extracted features through the convolutional neural network, indicating femoral neck was the most common cause of femoral fractures across all three grading stages.

Conclusions: We developed a DL model for the multi-classification of osteoporosis using real-world clinical data, combining CT-scanned images with variables. This implies that DL can be fully applied to medical data for the classification stage of osteoporosis. Our results suggest that abdominal CT could be important in osteoporosis screening and could lead to appropriate treatment for the reduction of osteoporotic fractures.

Artificial intelligence

Bone density

Clinical Decision support systems

Deep learning

Osteoporosis

Osteoporosis has emerged as a growing global health concern exacerbated by an aging population and longer life spans. [1] It is a systemic skeletal disease that decreases bone density and weakens the bone microarchitecture, increasing the risk of fractures. [2] Furthermore, it often remains asymptomatic until a fracture occurs, creating significant challenges for early diagnosis. This delay in detection can result in heightened risks of fractures, diminished quality of life, and elevated mortality rates. [3, 4] Among the older population aged > 70 years, the incidence of the disease is 18% in males and 68.5% in females, and the prevalence of osteoporosis appears to increase rapidly after menopause in females aged > 50 years. [3, 5, 6]

Dual-energy X-ray absorptiometry (DXA) is a commonly employed and established method for diagnosing osteoporosis, which delivers highly accurate measurements of bone mineral density (BMD) while minimizing radiation exposure. [7] Its ability to focus on key areas, such as the hip and spine, makes it invaluable for assessing the risk of fracture in clinical settings. However, it may not provide a comprehensive view of bone health. [8, 9] This limitation combined with the low screening rates highlights the impracticality of relying solely on DXA for early diagnosis. [9–12] Therefore, there is a need to develop additional methods for detecting osteoporosis. One possible way to improve osteoporosis identification rates is to use bone data that is obtained from abdominal computed tomography (CT) scans performed for other indications. [12–16] Patients who undergo abdominal CT have a potential opportunity for BMD screening of the femur without the need for any additional imaging, radiation exposure, or patient time. [13] Previous research has demonstrated the feasibility of predicting osteoporosis by examining the femoral region in abdominal CT scans. [12, 17, 18] Therefore, abdominal CT is considered a valuable tool for assessing the risk of osteoporosis and osteopenia as well as for distinguishing different stages of the disease.

With an exponential increase in computing power in the era of big data, deep-learning (DL) approaches have been rapidly adopted for the diagnosis of bone diseases including osteoporosis. [19, 20] Compared with traditional clinical tools, artificial intelligence (AI)-based methods offer higher accuracy by analyzing the complex relationships between the diverse features of osteoporosis. [2, 21–23]

Therefore, we developed a deep learning (DL) model to classify osteoporosis into three grading stages, as well as conducted gradient-weighted class activation mapping (Grad-CAM) to provide a clinical interpretation of the results. We also compared the performance of abdominal CT with information from BMD and further investigated the effect of combining the two on performance.

Patient characteristics

Table 1. Patients characteristics.

	Normal (N=1,068)	Osteopenia (N=925)	Osteoporosis (N=1,019)	Total (N=3,012)
Age^**	36.6 ± 15.6	62.5 ± 11.9	76.5 ± 9.6	58.1 ± 21.1
Sex^*
Male	291 (27.2%)	134 (14.5%)	220 (21.6%)	645 (21.4%)
Female	777 (72.8%)	791 (85.5%)	799 (78.4%)	2367 (78.6%)
Height^*	1.6 ± 0.1	1.6 ± 0.1	1.5 ± 0.1	1.6 ± 0.1
Weight^*	64.2 ± 13.6	59.7 ± 9.7	53.7 ± 9.1	59.3 ± 11.9
BMI^*	24.1 ± 4.4	24.5 ± 3.7	23.0 ± 3.7	23.9 ± 4.0

The superscript * means that the variable has statistical significance (P < 0.001) and ** means that the variable has significance (P < 0.05) among the three groups by Welch's ANOVA and Chi-squared. As a result of post hoc analysis of individual groups, statistically significant differences showed between all individual groups. However, there was no statistic significant in BMI between normal and osteopenia and sex between normal and osteoporosis.

Values are presented as mean ± SD unless otherwise stated. BMI; Body Mass Index

The characteristics of all the patients included in this study are presented in Table 1. The mean age across all patients was 58.1 years, with normal individuals averaging 36.6 years, those with osteopenia at 62.5 years, and those with osteoporosis at 76.5 years. Among all patients, 2367 were females (78.6%), of whom 777 (72.8%) were normal, 791 (85.5%) had osteopenia, and 799 (78.4%) had osteoporosis. Notably, there was a trend indicating a decrease in patient weight as the disease severity increased; the mean weight for normal patients was 64.2 kg, for those with osteopenia it was 59.7 kg, and for those with osteoporosis, it was 53.7 kg. The differences among the three groups for all variables were P < 0.001, as determined by ANOVA. As a result of post hoc analysis of individual groups, statistically significant differences showed between all individual groups, except in BMI between the normal and osteopenia groups and sex between the normal and osteoporosis groups.

Model performance

Table 2. Model performance.

Model	Data modality	ACC	Sensitivity	Specificity	Precision	F1-Score	AUC
6-layer CNN	Image	0.7944	0.7951	0.8971	0.8006	0.7978	0.9326
2-layer FNN	Clinical variables	0.6892	0.69	0.8444	0.6922	0.6911	0.8591
6-layer CNN with 2-layer FNN	Multi-modality	0.8018	0.8021	0.9006	0.8043	0.8032	0.9385

The dataset total counts of 3,012. Train data counts of 2,419 and test data counts of 593. ACC, Accuracy; AUC, Area under the receiver operating characteristic curve; CNN, Convolutional neural network; FNN, Fully connected neural network;

Table 2 summarizes the model performance results. All result metrics in the multimodal dataset showed that the highest performance was more significant than the independent use of demographic information for grading stage classification. The results of the multimodal dataset models showed the highest with an area under the receiver operating characteristic curve (AUC) of 0.94 and accuracy (ACC) of 0.80. This is more than 10% better than a model that uses demographic information. The model using demographic information independently shows the worst performance with AUC of 0.85 and ACC of 0.68. In addition, there was a larger difference in sensitivity than in specificity between the two datasets. The model using a multi-modal dataset had a sensitivity and specificity of 0.80 and 0.90, respectively, whereas the independent use of demographic data had a sensitivity and specificity of 0.69 and 0.84, respectively. Similar results were observed for the precision and recall.

Identification of informative features for classification

To identify informative features extracted using the convolutional neural network (CNN) model, Grad-CAM was used (Figure 4). This allowed for the visualization of evidence during the analysis process. In this study, both the right and left femurs were tested depending on the site; therefore, a set limited to the left femur was used to compare the shape and tilt position of the femur more accurately. Figure 4 is a visualization of images and gradients extracted from the evaluation data set, averaged, and transparency weighted at 0.7 and 0.3, respectively. The parts that contribute significantly to the model results are shown in green and the femoral neck can be seen to be activated at all stages.

In this study, we conducted a multi-classification of osteoporosis into the following grading stages: normal, osteopenia, and osteoporosis through DL using abdominal CT, which is already widely used in clinical practice. Using this DL model, CT images captured for diverse medical purposes can be used to screen latent patients at risk of osteoporosis without additional costs and radiation exposure. Here, modeling was performed separately using CT images, demographic data, and a combination of the two types of data. The model using multimodality data performed well for multi-classification with AUC and ACC of 0.94 and 0.80, respectively, by combining data from two different modalities.

In general, osteoporosis has a high prevalence and is known to present with difficulty in detecting fractures before disease progression. [24] Therefore, most older patients miss the opportunity to combat the risk of fracture owing to decreased bone density. Women > 65 and men > 70 years of age are exposed to many factors that can cause osteoporosis, such as low body weight and a history of previous fractures. [25] Although DXA is a representative method used to diagnose osteoporosis, it has some disadvantages, such as a low utility rate and radiation exposure. Moreover, practical challenges, such as the requirement for specialized equipment knowledge and limited accessibility due to low penetration rates exist. Consequently, an alternative approach to osteoporosis detection involves utilizing bone data derived from abdominal CT scans, a method widely recognized as a valuable tool for osteoporosis screening in general. [16, 22, 26, 27] Abdominal CT is a medical imaging modality that includes the spine and femur and is known to be very useful in accurately measuring the risk of osteoporosis in the area of interest. Proximal femoral fractures are fatal fractures with high morbidity and mortality rates despite representing only a small proportion of osteoporotic fractures.

Thus, our study offers several novelties for the multi-classification of osteoporosis into three stages using the femur region on abdominal CT images. First, our study results provide an opportunity to overcome the shortcomings of DXA and quickly respond to the potential risk of the disease using widely used and easily obtained CT images and demographic data. In particular, we utilized real-world clinical data to address the challenge of concurrently classifying individuals into the normal, osteopenia, and osteoporotic stages, leveraging observations from the femur region in abdominal CT. Previous studies have primarily focused on classifying individuals as either normal or having osteoporosis, with limited attention given to classifying osteopenia—a stage that falls between the two disease categories—as either normal or osteoporotic. Moreover, few studies have classified osteopenia, which lies between the two disease stages, as normal or osteoporosis (Table 3).

We obtained an AUC of 0.94, ACC of 0.81, precision and sensitivity of 0.8, and specificity and recall of 0.1, upon utilizing a multimodal dataset. These outcomes surpass those reported in several recent studies. [19, 28] Notably, our results outperform the performance achieved by machine learning-based X-ray image analysis in many existing studies. [23, 28, 29] Zhang et al. performed multi-classification using the same CNN model used by our team, but showed a performance of AUC 0.81 and ACC 0.6, lower than our results. [28] Liu et al. performed binary classification for each of the three groups (normal, osteopenia, osteoporosis) and presented results of AUC 0.88 for classification between normal and osteopenia, AUC 0.87 for classification between normal and osteoporosis, and AUC 0.75 for classification between osteopenia and osteoporosis. [29] Yamamoto et al. presented relatively high model performance results of AUC 0.93 and ACC 0.88 as a result of classification between normal and osteoporosis. [23] However, these studies used X-ray images and were conducted on a smaller number of patients than our study. In addition, Yasaka et al.’s study exhibited superior performance to ours in classifying normal and osteopenia; however, generalization would be difficult because the dataset for validation was very small in that study, [30] unlike that of ours. A similarity across all the studies, including ours, is the use of CT images. However, we can present high accessibility for use in clinical practice and enhance clinical applicability by performing multi-class classifications. Second, our study presents the explanatory potential of DL models.

One prevalent issue with DL-based approaches is the ‘black box’ nature of these models, which hinders a clear understanding of their internal processes. [31] Misinterpretations by AI-based models can lead to incorrect diagnoses, emphasizing the need for model validation. In our study, we employed Grad-CAM to visualize and elucidate the model's inferred rationale, confirming the alignment between the feature regions within the model and clinically relevant areas (Fig. 4). [32–34] As a result of our Grad-CAM analysis, we focused on feature extraction, with a specific emphasis on the femoral neck. This alignment with clinical knowledge underscores the femoral neck as the most vulnerable area to osteoporosis. Furthermore, our study carries implications for expanding its clinical applications. By amassing data spanning all age groups ranging from 20–70, we can facilitate multiple classifications of osteoporosis for various age brackets. Moreover, our findings remain applicable even for individuals with scoliosis or those who have undergone femur-related surgery, as we leverage the entire thigh region observable in abdominal CT scans. Additionally, by harnessing all femoral bone images within abdominal CT scans, categorized into the neck, head, and torso, we obtain input data with minimal need for preprocessing in most cases. This presents significant potential for the rapid proliferation of computer-aided diagnosis. Lastly, we conducted a comprehensive analysis to identify errors that may have arisen from differences in data distribution while classifying osteoporosis (Supplementary Fig. 1). It is reasonable to encounter classification challenges at the boundary points that demarcate these stages, given that disease risks can vary within the same stage based on T-score values. To visualize these discrepancies, we used the T-score. Our analysis reaffirmed the accuracy of most classifications between normal and osteoporotic tissues while highlighting that errors primarily concentrate on the transitional boundaries between disease stages.

However, this study has several limitations. First, the performance was guaranteed only for contrast-enhanced abdominal CT data. Although CT is a common imaging modality, it seldom provides BMD information in the clinic, owing to technical difficulties. Thus, DXA is required to measure BMD at the expense of additional radiation exposure. Furthermore, abdominal CT, which is generally used for patients who do not have kidney function abnormalities or contrast agent side effects, is a frequently scanned medical image. Accordingly, we conducted a study using contrast-enhanced abdominal CT which showed high accuracy in osteoporosis classification. Second, there was a time gap in the data collected in this study. All patient data were obtained using a concomitant CT scan within 3 months before or after the DXA scan to collect as much data as possible. Therefore, based on a single DXA examination, two or more CT images may be matched for the same patient. Third, data were collected from a single medical institution. It is difficult to prove this effect using data obtained from other institutions or CT equipment. In the future, we plan to collect more data from several machines and hospitals to reduce bias and increase robustness. Fourth, we confirmed that using clinical structured data and unstructured CT image data simultaneously improved the performance compared to using individual data independently. However, the results obtained using only images were not significantly different from those obtained using multimodal data. This finding suggests that the demographic data used in this study had only a minor effect. Improved performance can be expected by additionally using clinical variables directly related to bone density, such as drugs and disease history; however, this may result in a trade-off for widely used in clinical practice.

In conclusion, we developed a DL model for the multi-classification of osteoporosis using real-world clinical data combining CT scanned images with variables. Additionally, we discussed important features selected based on Grad-CAM technology. This implies that DL can be fully applied to medical data for the classification of osteoporosis. In addition, our results suggest that abdominal CT could be used as important data in osteoporosis screening and lead to appropriate treatment for the reduction of osteoporotic fractures.

Study Design and Participants

A total of 3,012 image data were collected from 2,126 patients who underwent contrast-enhanced abdominal CT and DXA between January 2015 and October 2021. All CT images were acquired within ± 3 months of the DXA scan, and data were labeled using T-Score from the DXA. We excluded patients who had foreign bodies resulting from femoral surgery, implanted artificial joints, or those lacking coronal CT phases from our study. Furthermore, since the majority of patients who underwent both DXA and CT scans within the same study period exhibited abnormal bone density, we decided to include additional data from patients in their twenties who had exclusively undergone contrast-enhanced abdominal CT scans under identical conditions. A qualified radiologist assessed these patients and excluded those who did not have a fracture or chronic disease, and included those with normal bone density. Consequently, we developed a model to evaluate patients' bone density risk by utilizing their CT images and demographic information, such as gender, age, height, weight, and body mass index (BMI). For a more comprehensive analysis, we also constructed a model that incorporated both types of data. The patient recruitment process is illustrated in Fig. 1.

Image acquisition and measurements

Figure 2 shows images for each step. [25] All enhanced CT images were acquired using dual-source CT scanners with 64 and 128 detectors (SOMATOM Definition and SOMATOM Definition Flash; Siemens Healthcare, Forchheim, Germany), with the following parameters: detector collimation, 1.2 mm or 0.6 mm; field of view, 50 cm; Tube voltage and tube current for x-ray exposure, 125 mA; 80–140 kVp; beam width, 38.4 mm or 28.8 mm; beam pitch, 0.6; and slice thickness 2–4 mm. All scanned images were obtained 80 s after the administration of the contrast agent at a rate of 2.6 mL/s using an auto-injector. Subsequently, 10 ml saline was injected at a rate of 2.5 mL/s. DXA was performed using Lunar Prodigy and Lunar Prodigy Advance (GE Healthcare Systems, USA), and the T-score was calculated as the standard deviation compared with the BMD of women aged 20–40 years. According to the World Health Organization definitions, patients are classified based on their DXA-derived T-Score as follows: normal (T-Score ≥ − 1.0), osteopenia (–2.5 < T-Score < − 1.0), and osteoporosis (T-Score ≤ − 2.5).

Image preprocessing

The acquisition of the phase, encompassing the femoral neck, head, and body, during contrast-enhanced abdominal CT was carried out following a thorough review by a radiologist. The selected digital imaging and communications in the medicine (DICOM) files were converted into grayscale images that could be used for training. The brightness and contrast of the images were adjusted using the specified window value parameters. By setting the window width and center parameters, the maximum and minimum values of the window can be defined as follows:

$${w}_{max}={w}_{center}+{(w}_{width}/2)$$

$${w}_{min}={w}_{center}-{(w}_{width}/2)$$

It can also be converted to the original unit of measure in the device generated using ${rescale}_{slope}$ and ${rescale}_{intercept}$.

$${rescale}_{pixel}={pixel\times rescale}_{slope}-{rescale}_{intercept}$$

In this study, by setting ${window}_{center}$ to 500 and ${window}_{width}$ to 2000, Eq. (4) was used to convert the bones into an image such that they stood out.

$$pixel=\frac{{rescale}_{pixel}-{window}_{min}}{{window}_{max}-{window}_{min}}\times 255$$

Because the images obtained all had different sizes depending on the characteristics of the patient, and many other parts besides the femur were included, it was necessary to obtain the regions of interest. All the images were cropped to 250 × 250 pixels to include the femoral neck and head. During the training, augmentation was performed on the preprocessed images to prevent overfitting. The training image, rotation, zoom in/out, and translation (up, down, left, and right) methods were randomly changed from − 10 to 10, and a vertical/horizontal flip method was applied. The probability of applying each augmentation method was set to 0.5.

Implementation models

We divided the training dataset into two subsets: training (80%) and validation (20%) datasets, and all hyperparameters were tuned during the validation phase. The validation set consisted of 20% of the training set. Our proposed model was divided into segments designed to learn from both images and demographic information. Specifically, we constructed a model dedicated to processing CT images and demographic data, while also conducting a comprehensive analysis by combining both datasets. In addition, we conducted a multimodal data analysis by combining both sets of data. As depicted in Fig. 3, the model architecture for learning image features consists of six convolutional layers and two fully connected layers, whereas the model for learning demographic information consists of three fully connected layers. Each modality is learned and transformed into a feature map, which is the key to the classification. The two generated feature maps were expressed by concatenation and classified into normal, osteopenia, and osteoporosis groups through two fully connected layers. Each layer uses a rectified linear unit (ReLU) activation function and is configured to learn quickly using batch normalization. The Adam optimizer was used, the mini-batch size was set to 32, and the epoch was set to 500. To shorten the experimental time, if learning did not improve after 50 epochs, it was ended early. The initial learning rate was set to 0.001 and adjusted to 70% if the performance did not improve after 30 epochs using the scheduler. A deep learning model was developed using Python 3.8.10 version and Tensorflow 2.4.0.

Informative feature identification for multi-classification

To identify informative features extracted through CNN models, Grad-CAM was used. The feature map could be visualized with the average pixel value up to the final layers. We identified regions in the femur by applying the ReLU activation function to visualize important parts of the model during the analysis process.

Statistical analysis

The Chi-squared (χ2) test and analysis of variance (ANOVA) were used to confirm the differences in ratio and mean among the three groups (normal, osteopenia, osteoporosis). After conducting Levene's test to check the equality of variances, an analysis of variance (ANOVA) was performed to test the difference in the means among the three groups. In addition, if the assumption of equal variance was not satisfied, Welch's ANOVA was performed to test for mean differences among groups, and the Games-Howell test was conducted to confirm post-hoc analysis.

Ethics Statement

This study was approved by the Institutional Review Board (IRB) (IRB number: A-2021-03-020) and the requirement for informed consent was waived because of the non-interventional observational nature of the study.

AI: Artificial intelligence

ANOVA: Analysis of variance

AUC: Area under the receiver operating characteristic curve

BMD: Bone mineral density

CNN: Convolutional neural network

CT: Computed tomography

DICOM: Digital imaging and communications in the medicine.

DL: Deep learning

DXA: Dual-energy X-ray absorptiometry

Grad-CAM: Gradient-Weighted Class Activation Mapping

ReLU: Rectified linear unit

Authors’ Contributions

TJH collected and analyzed the data, curation of data, visualized the results, and wrote the manuscript.

HSK collected and analyzed the data. DHH, NYY and SUK investigated the materials. WJK, HSC, JHK and KWM revised and reviewed the manuscript. SWP wrote and supervised manuscript, funding acquisition, analyzed data and interpreted the results. SHB designed the study conceptualization, collected and analyzed data, interpreted the results, wrote and revised the manuscript, and funding acquisition. All authors approved for the final manuscript.

Ethics approval and consent to participate

Consent for publication

Not applicable.

Competing interests

All authors declare that they have no competing interests.

Availability of data and materials

The datasets analyzed in the current study are available from the corresponding author on reasonable request because it is not publicly available data.

Funding

This research was supported by "Regional Innovation Strategy (RIS)" through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (MOE)(2022RIS-005) and supported by the Promotion of Innovative Businesses for Regulation-Free Special Zones funded by the Ministry of Small and Medium-sized Enterprises (SMEs) and Startups (MSS, South Korea) (P0020626). The Convergence security core talent training business (Gangwon University)(1711193144) by the Ministry of Science and ICT is also supported for this study.

Sözen T, Özışık L, Başaran NÇ. An overview and management of osteoporosis. Eur J Rheumatol. 2017;4:46–56. https://doi.org/10.5152/eurjrheum.2016.048
Abbouchie H, Raju N, Lamanna A, Chiang C, Kutaiba N. Screening for osteoporosis using L1 vertebral density on abdominal CT in an Australian population. Clin Radiol. 2022;77:e540–8. https://doi.org/10.1016/j.crad.2022.04.002
NIH Consensus Development Panel on Osteoporosis Prevention, Diagnosis, and Therapy. Osteoporosis prevention, diagnosis, and therapy. JAMA. 2001;285:785–95. https://doi.org/10.1001/jama.285.6.785
Löffler MT, Jacob A, Scharr A, Sollmann N, Burian E, El Husseini M, et al. Automatic opportunistic osteoporosis screening in routine CT: Improved prediction of patients with prevalent vertebral fractures compared to DXA. Eur Radiol. 2021;31:6069–77. https://doi.org/10.1007/s00330-020-07655-2
Expert Panel on Musculoskeletal Imaging, Yu JS, Krishna NG, Fox MG, Blankenbaker DG, Frick MA, et al. ACR appropriateness Criteria® osteoporosis and bone mineral density: 2022 update. J Am Coll Radiol. 2022;19:S417–32. https://doi.org/10.1016/j.jacr.2022.09.007
Ahn SH, Park SM, Park SY, Yoo JI, Jung HS, Nho JH, et al. Osteoporosis and osteoporotic fracture fact sheet in Korea. J Bone Metab. 2020;27:281–90. https://doi.org/10.11005/jbm.2020.27.4.281
Dimai HP. Use of dual-energy X-ray absorptiometry (DXA) for diagnosis and fracture risk assessment; WHO-criteria, T- and Z-score, and reference databases. Bone. 2017;104:39–43. https://doi.org/10.1016/j.bone.2016.12.016
King AB, Fiorentino DM. Medicare payment cuts for osteoporosis testing reduced use despite tests’ benefit in reducing fractures. Health Aff (Millwood). Accessed Sep 9, 2023. https://www.healthaffairs.org/doi/full/10.1377/hlthaff.2011.0233. 2011;30:2362–70. https://doi.org/10.1377/hlthaff.2011.0233
Incidence of fractures of the femur, including subtrochanteric, up to 8 years since initiation of oral bisphosphonate therapy: A register-based cohort study using the US MarketScan claims databases. Accessed Sep 9, 2023. https://link.springer.com/article/10.1007/s00198-012-1952-7. SpringerLink
Amarnath ALD, Franks P, Robbins JA, Xing G, Fenton JJ. Underuse and overuse of osteoporosis screening in a regional health system: A retrospective cohort study. J Gen Intern Med. 2015;30:1733–40. https://doi.org/10.1007/s11606-015-3349-8
Curtis JR, Carbone L, Cheng H, Hayes B, Laster A, Matthews R, et al. Longitudinal trends in use of bone mass measurement among older Americans, 1999–2005. J Bone Miner Res. 2008;23:1061–7. https://doi.org/10.1359/JBMR.080232
Lee SJ, Anderson PA, Pickhardt PJ. Predicting future hip fractures on routine abdominal CT using opportunistic osteoporosis screening measures: A matched case-control study. AJR Am J Roentgenol. 2017;209:395–402. https://doi.org/10.2214/AJR.17.17820
Summers RM, Baecher N, Yao J, Liu J, Pickhardt PJ, Choi JR, et al. Feasibility of simultaneous CT colonography and fully-automated bone mineral densitometry in a single examination. J Comput Assist Tomogr. 2011;35:212–6. https://doi.org/10.1097/RCT.0b013e3182032537
Pickhardt PJ, Lee LJ, del Rio AM, Lauder T, Bruce RJ, Summers RM, et al. Simultaneous screening for osteoporosis at CT colonography: Bone mineral density assessment using MDCT attenuation techniques compared with the DXA reference standard. J Bone Miner Res. 2011;26:2194–203. https://doi.org/10.1002/jbmr.428, PMID: 21590738, PMCID: PMC3304444
Pickhardt PJ, Pooler BD, Lauder T, Del Rio AM, Bruce RJ, Binkley N. Opportunistic screening for osteoporosis using abdominal computed tomography scans obtained for other indications. Ann Intern Med. 2013;158:588–95. https://doi.org/10.7326/0003-4819-158-8-201304160-00003
Lee SJ, Binkley N, Lubner MG, Bruce RJ, Ziemlewicz TJ, Pickhardt PJ. Opportunistic screening for osteoporosis using the sagittal reconstruction from routine abdominal CT for combined assessment of vertebral fractures and density. Osteoporos Int. 2016;27:1131–6. https://doi.org/10.1007/s00198-015-3318-4
Rosen MP, Siewert B, Sands DZ, Bromberg R, Edlow J, Raptopoulos V. Value of abdominal CT in the emergency department for patients with abdominal pain. Eur Radiol. 2003;13:418–24. https://doi.org/10.1007/s00330-002-1715-5
Effect of IV contrast on lumbar trabecular attenuation at routine abdominal CT: Correlation with DXA and implications for opportunistic osteoporosis screening. Accessed Sep 9, 2023. https://link.springer.com/article/10.1007/s00198-015-3224-9. SpringerLink
Suh B, Yu H, Kim H, Lee S, Kong S, Kim JW, et al. Interpretable deep-learning approaches for osteoporosis risk screening and individualized feature analysis using large population-based data: Model development and performance evaluation. J Med Internet Res. 2023;25:e40179. https://doi.org/10.2196/40179
LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521:436–44. https://doi.org/10.1038/nature14539
Tarekegn A, Ricceri F, Costa G, Ferracin E, Giacobini M. Predictive modeling for frailty conditions in elderly people: Machine learning approaches. JMIR Med Inform. 2020;8:e16678. https://doi.org/10.2196/16678
Smets J, Shevroja E, Hügle T, Leslie WD, Hans D. Machine learning solutions for osteoporosis—A review. J Bone Miner Res. 2021;36:833–51. https://doi.org/10.1002/jbmr.4292
Yamamoto N, Sukegawa S, Kitamura A, Goto R, Noda T, Nakano K, et al. Deep learning for osteoporosis classification using hip radiographs and patient clinical covariates. Biomolecules. 2020;10:1534. https://doi.org/10.3390/biom10111534
Zeytinoglu M, Jain RK, Vokes TJ. Vertebral fracture assessment: Enhancing the diagnosis, prevention, and treatment of osteoporosis. Bone. 2017;104:54–65. https://doi.org/10.1016/j.bone.2017.03.004. Epub Mar 8 2017. PMID: 28285014
World Health Organization. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: Report of a WHO study group [meeting held in Rome from 22 to 25 June 1992]. World Health Organization; 1994
Yang G, Wang H, Wu Z, Shi Y, Zhao Y. Prediction of osteoporosis and osteopenia by routine computed tomography of the lumbar spine in different regions of interest. J Orthop Surg Res. 2022;17:454. https://doi.org/10.1186/s13018-022-03348-2
Kim YW, Kim JH, Yoon SH, Lee JH, Lee CH, Shin CS, et al. Vertebral bone attenuation on low-dose chest CT: Quantitative volumetric analysis for bone fragility assessment. Osteoporos Int. 2017;28:329–38. https://doi.org/10.1007/s00198-016-3724-2
Zhang B, Yu K, Ning Z, Wang K, Dong Y, Liu X, et al. Deep learning of lumbar spine X-ray for osteopenia and osteoporosis screening: A multicenter retrospective cohort study. Bone. 2020;140:115561. https://doi.org/10.1016/j.bone.2020.115561
Liu J, Wang J, Ruan W, Lin C, Chen D. Diagnostic and gradation model of osteoporosis based on improved deep U-net network. J Med Syst. 2019;44:15. https://doi.org/10.1007/s10916-019-1502-3
Yasaka K, Akai H, Kunimatsu A, Kiryu S, Abe O. Prediction of bone mineral density from computed tomography: Application of deep learning with a convolutional neural network. Eur Radiol. 2020;30:3549–57. https://doi.org/10.1007/s00330-020-06677-0
Lipton ZC. The Mythos of Model Interpretability: In machine learning, the concept of interpretability is both important and slippery. Queue. 2018;16:31–57. https://doi.org/10.1145/3236386.3241340
Jones G, Nguyen T, Sambrook P, Kelly PJ, Eisman JA. Progressive loss of bone in the femoral neck in elderly people: Longitudinal findings from the Dubbo osteoporosis epidemiology study. BMJ. 1994;309:691–5. https://doi.org/10.1136/bmj.309.6956.691
Wright NC, Looker AC, Saag KG, Curtis JR, Delzell ES, Randall S, et al. The recent prevalence of osteoporosis and low bone mass in the United States based on bone mineral density at the femoral neck or lumbar spine. J Bone Miner Res. 2014;29:2520–6. https://doi.org/10.1002/jbmr.2269
Kanis JA. Diagnosis of osteoporosis and assessment of fracture risk. Lancet. 2002;359:1929–36. https://doi.org/10.1016/S0140-6736(02)08761-5

No competing interests reported.

230924SupplementaryFigure.pptx

Download PDF

Version 1

posted

You are reading this latest preprint version

Multi-classification of Grading Stages for Osteoporosis Using Abdominal Computed Tomography with Clinical Variables: Application of Deep Learning with a Convolutional Neural Network

Status:

Version 1

Abstract

Figures

BACKGROUND

RESULTS

DISCUSSION

CONCLUSIONS

METHODS

Study Design and Participants

Image acquisition and measurements

Image preprocessing

Implementation models

Informative feature identification for multi-classification

Statistical analysis

Ethics Statement

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1