Deep-learning-based image quality enhancement of CT-like MR imaging in patients with suspected traumatic shoulder injury

PURPOSE

To evaluate the diagnostic performance of CT-like MR images reconstructed with an algorithm combining compressed sense (CS) with deep learning (DL) in patients with suspected osseous shoulder injury compared to conventional CS-reconstructed images.

METHODS

Thirty-two patients (12 women, mean age 46 ± 14.9 years) with suspected traumatic shoulder injury were prospectively enrolled into the study. All patients received MR imaging of the shoulder, including a CT-like 3D T1-weighted gradient-echo (T1 GRE) sequence and in case of suspected fracture a conventional CT. An automated DL-based algorithm, combining CS and DL (CS DL) was used to reconstruct images of the same k-space data as used for CS reconstructions. Two musculoskeletal radiologists assessed the images for osseous pathologies, image quality and visibility of anatomical landmarks using a 5-point Likert scale. Moreover, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated.

RESULTS

Compared to CT, all acute fractures (n = 23) and osseous pathologies were detected accurately on the CS only and CS DL images with almost perfect agreement between the CS DL and CS only images (κ 0.95 (95 %confidence interval 0.82-1.00). Image quality as well as the visibility of the fracture lines, bone fragments and glenoid borders were overall rated significantly higher for the CS DL reconstructions than the CS only images (CS DL range 3.7-4.9 and CS only range 3.2-3.8, P = 0.01-0.04). Significantly higher SNR and CNR values were observed for the CS DL reconstructions (P = 0.02-0.03).

CONCLUSION

Evaluation of traumatic shoulder pathologies is feasible using a DL-based algorithm for reconstruction of high-resolution CT-like MR imaging.

Assessment of glenoid bone loss and other osseous shoulder pathologies comparing MR-based CT-like images with conventional CT

OBJECTIVES

To evaluate and compare the diagnostic performance of CT-like images based on a 3D T1-weighted spoiled gradient-echo sequence (T1 GRE), an ultra-short echo time sequence (UTE), and a 3D T1-weighted spoiled multi-echo gradient-echo sequence (FRACTURE) with conventional CT in patients with suspected osseous shoulder pathologies.

MATERIALS AND METHODS

Patients with suspected traumatic dislocation of the shoulder (n = 46, mean age 40 ± 14.5 years, 19 women) were prospectively recruited and received 3-T MR imaging including 3D T1 GRE, UTE, and 3D FRACTURE sequences. CT was performed in patients with acute fractures and served as standard of reference (n = 25). Agreement of morphological features between the modalities was analyzed including the glenoid bone loss, Hill-Sachs interval, glenoid track, and the anterior straight-line length. Agreement between the modalities was assessed using Bland-Altman plots, Student's t-test, and Pearson's correlation coefficient. Inter- and intrareader assessment was evaluated with weighted Cohen's κ and intraclass correlation coefficient.

RESULTS

All osseous pathologies were detected accurately on all three CT-like sequences (n = 25, κ = 1.00). No significant difference in the percentage of glenoid bone loss was found between CT (mean ± standard deviation, 20.3% ± 8.0) and CT-like MR images (FRACTURE 20.6% ± 7.9, T1 GRE 20.4% ± 7.6, UTE 20.3% ± 7.7, p > 0.05). When comparing the different measurements on CT-like images, measurements performed using the UTE images correlated best with CT.

CONCLUSION

Assessment of bony Bankart lesions and other osseous pathologies was feasible and accurate using CT-like images based on 3-T MRI compared with conventional CT. Compared to the T1 GRE and FRACTURE sequence, the UTE measurements correlated best with CT.

CLINICAL RELEVANCE STATEMENT

In an acute trauma setting, CT-like images based on a T1 GRE, UTE, or FRACTURE sequence might be a useful alternative to conventional CT scan sparing associated costs as well as radiation exposure.

KEY POINTS

• No significant differences were found for the assessment of the glenoid bone loss when comparing measurements of CT-like MR images with measurements of conventional CT images. • Compared to the T1 GRE and FRACTURE sequence, the UTE measurements correlated best with CT whereas the FRACTURE sequence appeared to be the most robust regarding motion artifacts. • The T1 GRE sequence had the highest resolution with high bone contrast and detailed delineation of even small fractures but was more susceptible to motion artifacts.

Evaluation of 3D MRI for early detection of bone edema associated with apical periodontitis

OBJECTIVES

To detect and evaluate early signs of apical periodontitis using MRI based on a 3D short-tau-inversion-recovery (STIR) sequence compared to conventional panoramic radiography (OPT) and periapical radiographs in patients with apical periodontitis.

MATERIALS AND METHODS

Patients with clinical evidence of periodontal disease were enrolled prospectively and received OPT as well as MRI of the viscerocranium including a 3D-STIR sequence. The MRI sequences were assessed for the occurrence and extent of bone changes associated with apical periodontitis including bone edema, periradicular cysts, and dental granulomas. OPTs and intraoral periapical radiographs, if available, were assessed for corresponding periapical radiolucencies using the periapical index (PAI).

RESULTS

In total, 232 teeth of 37 patients (mean age 62±13.9 years, 18 women) were assessed. In 69 cases reactive bone edema was detected on MRI with corresponding radiolucency according to OPT. In 105 cases edema was detected without corresponding radiolucency on OPT. The overall extent of edema measured on MRI was significantly larger compared to the radiolucency on OPT (mean: STIR 2.4±1.4 mm, dental radiograph 1.3±1.2 mm, OPT 0.8±1.1 mm, P=0.01). The overall PAI score was significantly higher on MRI compared to OPT (mean PAI: STIR 1.9±0.7, dental radiograph 1.3±0.5, OPT 1.2±0.7, P=0.02).

CONCLUSION

Early detection and assessment of bone changes of apical periodontitis using MRI was feasible while the extent of bone edema measured on MRI exceeded the radiolucencies measured on OPT.

CLINICAL RELEVANCE

In clinical routine, dental MRI might be useful for early detection and assessment of apical periodontitis before irreversible bone loss is detected on conventional panoramic and intraoral periapical radiographs.

X-ray dark-field chest radiography: a reader study to evaluate the diagnostic quality of attenuation chest X-rays from a dual-contrast scanning prototype

OBJECTIVES

To compare the visibility of anatomical structures and overall quality of the attenuation images obtained with a dark-field X-ray radiography prototype with those from a commercial radiography system.

METHODS

Each of the 65 patients recruited for this study obtained a thorax radiograph at the prototype and a reference radiograph at the commercial system. Five radiologists independently assessed the visibility of anatomical structures, the level of motion artifacts, and the overall image quality of all attenuation images on a five-point scale, with 5 points being the highest rating. The average scores were compared between the two image types. The differences were evaluated using an area under the curve (AUC) based z-test with a significance level of p ≤ 0.05. To assess the variability among the images, the distributions of the average scores per image were compared between the systems.

RESULTS

The overall image quality was rated high for both devices, 4.2 for the prototype and 4.6 for the commercial system. The rating scores varied only slightly between both image types, especially for structures relevant to lung assessment, where the images from the commercial system were graded slightly higher. The differences were statistically significant for all criteria except for the bronchial structures, the cardiophrenic recess, and the carina.

CONCLUSIONS

The attenuation images acquired with the prototype were assigned a high diagnostic quality despite a lower resolution and the presence of motion artifacts. Thus, the attenuation-based radiographs from the prototype can be used for diagnosis, eliminating the need for an additional conventional radiograph.

KEY POINTS

• Despite a low tube voltage (70 kVp) and comparably long acquisition time, the attenuation images from the dark-field chest radiography system achieved diagnostic quality for lung assessment. • Commercial chest radiographs obtained a mean rating score regarding their diagnostic quality of 4.6 out of 5, and the grating-based images had a slightly lower mean rating score of 4.2 out of 5. • The difference in rating scores for anatomical structures relevant to lung assessment is below 5%.

Evaluation of a deep learning-based reconstruction method for denoising and image enhancement of shoulder MRI in patients with shoulder pain

OBJECTIVES

To evaluate the diagnostic performance of an automated reconstruction algorithm combining MR imaging acquired using compressed SENSE (CS) with deep learning (DL) in order to reconstruct denoised high-quality images from undersampled MR images in patients with shoulder pain.

METHODS

Prospectively, thirty-eight patients (14 women, mean age 40.0 ± 15.2 years) with shoulder pain underwent morphological MRI using a pseudo-random, density-weighted k-space scheme with an acceleration factor of 2.5 using CS only. An automated DL-based algorithm (CS DL) was used to create reconstructions of the same k-space data as used for CS reconstructions. Images were analyzed by two radiologists and assessed for pathologies, image quality, and visibility of anatomical landmarks using a 4-point Likert scale.

RESULTS

Overall agreement for the detection of pathologies between the CS DL reconstructions and CS images was substantial to almost perfect (κ 0.95 (95% confidence interval 0.82-1.00)). Image quality and the visibility of the rotator cuff, articular cartilage, and axillary recess were overall rated significantly higher for CS DL images compared to CS (p < 0.03). Contrast-to-noise ratios were significantly higher for cartilage/fluid (CS DL 198 ± 24.3, CS 130 ± 32.2, p = 0.02) and ligament/fluid (CS DL 184 ± 17.3, CS 141 ± 23.5, p = 0.03) and SNR values were significantly higher for ligaments and muscle of the CS DL reconstructions (p < 0.04).

CONCLUSION

Evaluation of shoulder pathologies was feasible using a DL-based algorithm for MRI reconstruction and denoising. In clinical routine, CS DL may be beneficial in particular for reducing image noise and may be useful for the detection and better discrimination of discrete pathologies. Assessment of shoulder pathologies was feasible with improved image quality as well as higher SNR using a compressed sensing deep learning-based framework for image reconstructions and denoising.

KEY POINTS

• Automated deep learning-based reconstructions showed a significant increase in signal-to-noise ratio and contrast-to-noise ratio (p < 0.04) with only a slight increase of reconstruction time of 40 s compared to CS. • All pathologies were accurately detected with no loss of diagnostic information or prolongation of the scan time. • Significant improvements of the image quality as well as the visibility of the rotator cuff, articular cartilage, and axillary recess were detected.

Imaging of traumatic mandibular fractures in young adults using CT-like MRI: a feasibility study

OBJECTIVES

To assess and compare the diagnostic performance of CT-like images based on a three- dimensional (3D) T1-weighted spoiled gradient-echo sequence (3D T1 GRE) with CT in patients with acute traumatic fractures of the mandible.

MATERIALS AND METHODS

Subjects with acute mandibular fractures diagnosed on conventional CT were prospectively recruited and received an additional 3 T MRI with a CT-like 3D T1 GRE sequence. The images were assessed by two radiologists with regard to fracture localization, degree of dislocation, and number of fragments. Bone to soft tissue contrast, diagnostic confidence, artifacts, and overall image quality were rated using a five-point Likert-scale. Agreement of measurements was assessed using an independent t-test.

RESULTS

Fourteen subjects and 22 fracture sites were included (26 ± 3.9 years; 4 females, 10 males). All traumatic fractures were accurately detected on CT-like MRI (n = 22, κ 1.00 (95% CI 1.00-1.00)). There was no statistically significant difference in the assessment of the fracture dislocation (axial mean difference (MD) 0.06 mm, p = 0.93, coronal MD, 0.08 mm, p = 0.89 and sagittal MD, 0.04 mm, p = 0.96). The agreement for the fracture classification as well as the inter- and intra-rater agreement was excellent (range κ 0.92-0.98 (95% CI 0.96-0.99)).

CONCLUSION

Assessment of mandibular fractures was feasible and accurate using CT-like MRI based on a 3D T1 GRE sequence and is comparable to conventional CT.

CLINICAL RELEVANCE

For the assessment of acute mandibular fractures, CT-like MRI might become a useful alternative to CT in order to reduce radiation exposure particularly in young patients.

Associations between Bone Mineral Density and Longitudinal Changes of Vertebral Bone Marrow and Paraspinal Muscle Composition Assessed Using MR-Based Proton Density Fat Fraction and T2* Maps in Patients with and without Osteoporosis

Background: Proton-density fat fraction (PDFF) and T2* of the vertebrae, as well as the cross-sectional area (CSA) of the paraspinal musculature (PSM), have been suggested as biomarkers for bone fragility. The aim of this study was to longitudinally assess changes in PDFF, T2* and CSA of the PSM over 6 months in patients with and without osteoporosis. Methods: Opportunistic bone mineral density (BMD) measurements (BMD < 120 mg/cm3) were obtained from a CT acquired during the clinical routine work up in osteoporotic/osteopenic patients (n = 29, mean age 72.37 ± 10.12 years, 16 women). These patients were frequency-matched for age and sex to subjects with normal BMD values (n = 29). All study patients underwent 3T MR imaging at baseline and 6-month follow up, including spoiled gradient echo sequences for chemical shift encoding-based water-fat separation, from which T2* and PDFF values of the lumbar spine and the PSM were obtained. Moreover, the CSA of the PSM was assessed longitudinally. Changes in T2*, PDFF and CSA over 6 months were calculated for the vertebrae and PSM and associations with baseline BMD values were assessed. Results: The change in CSA of the PSM over 6 months was significantly lower in the osteoporotic/osteopenic group (−91.5 ± 311.7 mm2), compared to the non-osteoporotic group, in which the CSA increased (29.9 ± 164.0 mm2, p = 0.03). In a further analysis, patients with higher vertebral PDFF at baseline showed a significantly stronger increase in vertebral T2*, compared to those patients with lower vertebral PDFF at baseline (0.9 ± 1.6 ms vs. 0.0 ± 1.8 ms, p = 0.04). Moreover, patients with higher PSM PDFF at baseline showed a significantly stronger increase in vertebral T2*, compared to those patients with lower PSM PDFF at baseline (0.9 ± 2.0 ms vs. 0.0 ± 1.3 ms, p = 0.03). Conclusion: The PSM CSA decreased significantly longitudinally in patients with osteoporosis/osteopenia, compared to those without. Additionally, higher vertebral and PSM PDFF at baseline were associated with stronger changes in vertebral bone marrow T2*. Therefore, longitudinal PDFF and T2* mapping may be useful quantitative radiation-free tools for the assessment and prediction of muscle and bone health in patients with suspected osteoporosis/osteopenia.

Associations of Texture Features of Proton Density Fat Fraction Maps between Lumbar Vertebral Bone Marrow and Paraspinal Musculature

Chemical shift encoding-based water−fat MRI (CSE-MRI)-derived proton density fat fraction (PDFF) has been used for non-invasive assessment of regional body fat distributions. More recently, texture analysis (TA) has been proposed to reveal even more detailed information about the vertebral or muscular composition beyond PDFF. The aim of this study was to investigate associations between vertebral bone marrow and paraspinal muscle texture features derived from CSE-MRI-based PDFF maps in a cohort of healthy subjects. In this study, 44 healthy subjects (13 males, 55 ± 30 years; 31 females, 39 ± 17 years) underwent 3T MRI including a six-echo three-dimensional (3D) spoiled gradient echo sequence used for CSE-MRI at the lumbar spine and the paraspinal musculature. The erector spinae muscles (ES), the psoas muscles (PS), and the vertebral bodies L1-4 (LS) were manually segmented. Mean PDFF values and texture features were extracted for each compartment. Features were compared between males and females using logistic regression analysis adjusted for age and body mass index (BMI). All texture features of ES except for Sum Average were significantly (p < 0.05) different between men and women. The three global texture features (Variance, Skewness, Kurtosis) for PS as well as LS showed a significant difference between male and female subjects (p < 0.05). Mean PDFF measured in PS and ES was significantly higher in females, but no difference was found for the vertebral bone marrow’s PDFF. Partial correlation analysis between the texture features of the spine and the paraspinal muscles revealed a highly significant correlation for Variance(global) (r = 0.61 for ES, r = 0.62 for PS; p < 0.001 respectively). Texture analysis using PDFF maps based on CSE-MRI revealed differences between healthy male and female subjects. Global texture features in the lumbar vertebral bone marrow allowed for differentiation between men and women, when the overall PDFF was not significantly different, indicating that PDFF maps may contain detailed and subtle textural information beyond fat fraction. The observed significant correlation of Variance(global) suggests a metabolic interrelationship between vertebral bone marrow and the paraspinal muscles.

Evaluation of MR-derived simulated CT-like images and simulated radiographs compared to conventional radiography in patients with shoulder pain: a proof-of-concept study

Background

To evaluate the diagnostic value of MR-derived CT-like images and simulated radiographs compared with conventional radiographs in patients with suspected shoulder pathology.

Methods

3 T MRI of the shoulder including a 3D T1-weighted gradient echo sequence was performed in 25 patients (mean age 52.4 ± 18 years, 13 women) with suspected shoulder pathology. Subsequently a cone-beam forward projection algorithm was used to obtain intensity-inverted CT-like images and simulated radiographs. Two radiologists evaluated the simulated images separately and independently using the conventional radiographs as the standard of reference, including measurements of the image quality, acromiohumeral distance, critical shoulder angle, degenerative joint changes and the acromial type. Additionally, the CT-like MR images were evaluated for glenoid defects, subcortical cysts and calcifications. Agreement between the MR-derived simulated radiographs and conventional radiographs was calculated using Cohen’s Kappa.

Results

Measurements on simulated radiographs and conventional radiographs overall showed a substantial to almost perfect inter- and intra-rater agreement (κ = 0.69–1.00 and κ = 0.65–0.85, respectively). Image quality of the simulated radiographs was rated good to excellent (1.6 ± 0.7 and 1.8 ± 0.6, respectively) by the radiologists. A substantial agreement was found regarding diagnostically relevant features, assessed on Y- and anteroposterior projections (κ = 0.84 and κ = 0.69 for the measurement of the CSA; κ = 0.95 and κ = 0.60 for the measurement of the AHD; κ = 0.77 and κ = 0.77 for grading of the Samilson-Prieto classification; κ = 0.83 and κ = 0.67 for the grading of the Bigliani classification, respectively).

Conclusion

In this proof-of-concept study, clinically relevant features of the shoulder joint were assessed reliably using MR-derived CT-like images and simulated radiographs with an image quality equivalent to conventional radiographs. MR-derived CT-like images and simulated radiographs may provide useful diagnostic information while reducing the amount of radiation exposure.

MR-based proton density fat fraction (PDFF) of the vertebral bone marrow differentiates between patients with and without osteoporotic vertebral fractures

The bone marrow proton density fat fraction (PDFF) assessed with MRI enables the differentiation between osteoporotic/osteopenic patients with and without vertebral fractures. Therefore, PDFF may be a potentially useful biomarker for bone fragility assessment.

INTRODUCTION

To evaluate whether magnetic resonance imaging (MRI)-based proton density fat fraction (PDFF) of vertebral bone marrow can differentiate between osteoporotic/osteopenic patients with and without vertebral fractures.

METHODS

Of the 52 study patients, 32 presented with vertebral fractures of the lumbar spine (66.4 ± 14.4 years, 62.5% women; acute low-energy osteoporotic/osteopenic vertebral fractures, N = 25; acute high-energy traumatic vertebral fractures, N = 7). These patients were frequency matched for age and sex to patients without vertebral fractures (N = 20, 69.3 ± 10.1 years, 70.0% women). Trabecular bone mineral density (BMD) values were derived from quantitative computed tomography. Chemical shift encoding-based water-fat MRI of the lumbar spine was performed, and PDFF maps were calculated. Associations between fracture status and PDFF were assessed using multivariable linear regression models.

RESULTS

Over all patients, mean PDFF and trabecular BMD correlated significantly (r =  - 0.51, P < 0.001). In the osteoporotic/osteopenic group, those patients with osteoporotic/osteopenic fractures had a significantly higher PDFF than those without osteoporotic fractures after adjusting for age, sex, weight, height, and trabecular BMD (adjusted mean difference [95% confidence interval], 20.8% [10.4%, 30.7%]; P < 0.001), although trabecular BMD values showed no significant difference between the subgroups (P = 0.63). For the differentiation of patients with and without vertebral fractures in the osteoporotic/osteopenic subgroup using mean PDFF, an area under the receiver operating characteristic (ROC) curve (AUC) of 0.88 (P = 0.006) was assessed. When evaluating all patients with vertebral fractures, those with high-energy traumatic fractures had a significantly lower PDFF than those with low-energy osteoporotic/osteopenic vertebral fractures (P < 0.001).

CONCLUSION

MR-based PDFF enables the differentiation between osteoporotic/osteopenic patients with and without vertebral fractures, suggesting the use of PDFF as a potential biomarker for bone fragility.

Development and evaluation of machine learning models based on X-ray radiomics for the classification and differentiation of malignant and benign bone tumors

OBJECTIVES

To develop and validate machine learning models to distinguish between benign and malignant bone lesions and compare the performance to radiologists.

METHODS

In 880 patients (age 33.1 ± 19.4 years, 395 women) diagnosed with malignant (n = 213, 24.2%) or benign (n = 667, 75.8%) primary bone tumors, preoperative radiographs were obtained, and the diagnosis was established using histopathology. Data was split 70%/15%/15% for training, validation, and internal testing. Additionally, 96 patients from another institution were obtained for external testing. Machine learning models were developed and validated using radiomic features and demographic information. The performance of each model was evaluated on the test sets for accuracy, area under the curve (AUC) from receiver operating characteristics, sensitivity, and specificity. For comparison, the external test set was evaluated by two radiology residents and two radiologists who specialized in musculoskeletal tumor imaging.

RESULTS

The best machine learning model was based on an artificial neural network (ANN) combining both radiomic and demographic information achieving 80% and 75% accuracy at 75% and 90% sensitivity with 0.79 and 0.90 AUC on the internal and external test set, respectively. In comparison, the radiology residents achieved 71% and 65% accuracy at 61% and 35% sensitivity while the radiologists specialized in musculoskeletal tumor imaging achieved an 84% and 83% accuracy at 90% and 81% sensitivity, respectively.

CONCLUSIONS

An ANN combining radiomic features and demographic information showed the best performance in distinguishing between benign and malignant bone lesions. The model showed lower accuracy compared to specialized radiologists, while accuracy was higher or similar compared to residents.

KEY POINTS

• The developed machine learning model could differentiate benign from malignant bone tumors using radiography with an AUC of 0.90 on the external test set. • Machine learning models that used radiomic features or demographic information alone performed worse than those that used both radiomic features and demographic information as input, highlighting the importance of building comprehensive machine learning models. • An artificial neural network that combined both radiomic and demographic information achieved the best performance and its performance was compared to radiology readers on an external test set.

Associations of incidental vertebral fractures and longitudinal changes of MR-based proton density fat fraction and T2* measurements of vertebral bone marrow

BACKGROUND

Quantitative magnetic resonance imaging (MRI) techniques such as chemical shift encoding-based water-fat separation techniques (CSE-MRI) are increasingly applied as noninvasive biomarkers to assess the biochemical composition of vertebrae. This study aims to investigate the longitudinal change of proton density fat fraction (PDFF) and T2* derived from CSE-MRI of the thoracolumbar vertebral bone marrow in patients that develop incidental vertebral compression fractures (VCFs), and whether PDFF and T2* enable the prediction of an incidental VCF.

METHODS

In this study we included 48 patients with CT-derived bone mineral density (BMD) measurements at baseline. Patients that presented an incidental VCF at follow up (N=12, mean age 70.5 ± 7.4 years, 5 female) were compared to controls without incidental VCF at follow up (N=36, mean age 71.1 ± 8.6 years, 15 females). All patients underwent 3T MRI, containing a significant part of the thoracolumbar spine (Th11-L4), at baseline, 6-month and 12 month follow up, including a gradient echo sequence for chemical shift encoding-based water-fat separation, from which PDFF and T2* maps were obtained. Associations between changes in PDFF, T2* and BMD measurements over 12 months and the group (incidental VCF vs. no VCF) were assessed using multivariable regression models. Mixed-effect regression models were used to test if there is a difference in the rate of change in PDFF, T2* and BMD between patients with and without incidental VCF.

RESULTS

Prior to the occurrence of an incidental VCF, PDFF in vertebrae increased in the VCF group (Δ_PDFF=6.3 ± 3.1%) and was significantly higher than the change of PDFF in the group without VCF (Δ_PDFF=2.1 ± 2.5%, P=0.03). There was no significant change in T2* (Δ_T2*=1.7 ± 1.1ms vs. Δ_T2*=1.1 ± 1.3ms, P=0.31) and BMD (Δ_BMD=-1.2 ± 11.3mg/cm³ vs. Δ_BMD=-11.4 ± 24.1mg/cm³, P= 0.37) between the two groups over 12 months. At baseline, no significant differences were detected in the average PDFF, T2* and BMD of all measured vertebrae (Th11-L4) between the VCF group and the group without VCF (P=0.66, P=0.35 and P= 0.21, respectively). When assessing the differences in rates of change, there was a significant change in slope for PDFF (2.32 per 6 months, 95% confidence interval (CI) 0.31-4.32; P=0.03) but not for T2* (0.02 per 6 months, CI -0.98-0.95; P=0.90) or BMD (-4.84 per 6 months, CI -23.4-13.7; P=0.60).

CONCLUSIONS

In our study population, the average change of PDFF over 12 months is significantly higher in patients that develop incidental fractures at 12-month follow up compared to patients without incidental VCF, while T2* and BMD show no significant changes prior to the occurrence of the incidental vertebral fractures. Therefore, a longitudinal increase in bone marrow PDFF may be predictive for vertebral compression fractures.

Multitask Deep Learning for Segmentation and Classification of Primary Bone Tumors on Radiographs

Background An artificial intelligence model that assesses primary bone tumors on radiographs may assist in the diagnostic workflow. Purpose To develop a multitask deep learning (DL) model for simultaneous bounding box placement, segmentation, and classification of primary bone tumors on radiographs. Materials and Methods This retrospective study analyzed bone tumors on radiographs acquired prior to treatment and obtained from patient data from January 2000 to June 2020. Benign or malignant bone tumors were diagnosed in all patients by using the histopathologic findings as the reference standard. By using split-sample validation, 70% of the patients were assigned to the training set, 15% were assigned to the validation set, and 15% were assigned to the test set. The final performance was evaluated on an external test set by using geographic validation, with accuracy, sensitivity, specificity, and 95% CIs being used for classification, the intersection over union (IoU) being used for bounding box placements, and the Dice score being used for segmentations. Results Radiographs from 934 patients (mean age, 33 years ± 19 [standard deviation]; 419 women) were evaluated in the internal data set, which included 667 benign bone tumors and 267 malignant bone tumors. Six hundred fifty-four patients were in the training set, 140 were in the validation set, and 140 were in the test set. One hundred eleven patients were in the external test set. The multitask DL model achieved 80.2% (89 of 111; 95% CI: 72.8, 87.6) accuracy, 62.9% (22 of 35; 95% CI: 47, 79) sensitivity, and 88.2% (67 of 76; CI: 81, 96) specificity in the classification of bone tumors as malignant or benign. The model achieved an IoU of 0.52 ± 0.34 for bounding box placements and a mean Dice score of 0.60 ± 0.37 for segmentations. The model accuracy was higher than that of two radiologic residents (71.2% and 64.9%; P = .002 and P < .001, respectively) and was comparable with that of two musculoskeletal fellowship-trained radiologists (83.8% and 82.9%; P = .13 and P = .25, respectively) in classifying a tumor as malignant or benign. Conclusion The developed multitask deep learning model allowed for accurate and simultaneous bounding box placement, segmentation, and classification of primary bone tumors on radiographs. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Carrino in this issue.

Vertebral bone marrow T2* mapping using chemical shift encoding-based water-fat separation in the quantitative analysis of lumbar osteoporosis and osteoporotic fractures

BACKGROUND

Chemical shift encoding-based water-fat separation techniques have been used for fat quantification [proton density fat fraction (PDFF)], but they also enable the assessment of bone marrow T2*, which has previously been reported to be a potential biomarker for osteoporosis and may give insight into the cause of vertebral fractures (i.e., osteoporotic vs. traumatic) and the microstructure of the bone when applied to vertebral bone marrow.

METHODS

The 32 patients (78.1% with low-energy osteopenic/osteoporotic fractures, mean age 72.3±9.8 years, 76% women; 21.9% with high-energy traumatic fractures, 47.3±12.8 years, no women) were frequency-matched for age and sex to subjects without vertebral fractures (n=20). All study patients underwent 3T-MRI of the lumbar spine including sagittally acquired spoiled gradient echo sequences for chemical shift encoding-based water-fat separation, from which T2* values were obtained. Volumetric trabecular bone mineral density (BMD) and trabecular bone parameters describing the three-dimensional structural integrity of trabecular bone were derived from quantitative CT. Associations between T2* measurements, fracture status and trabecular bone parameters were assessed using multivariable linear regression models.

RESULTS

Mean T2* values of non fractured vertebrae in all patients showed a significant correlation with BMD (r=-0.65, P<0.001), trabecular number (TbN) (r=-0.56, P<0.001) and trabecular spacing (TbSp) (r=0.61, P<0.001); patients with low-energy osteoporotic vertebral fractures showed significantly higher mean T2* values than those with traumatic fractures (13.6±4.3 vs. 8.4±2.2 ms, P=0.01) as well as a significantly lower TbN (0.69±0.08 vs. 0.93±0.03 mm-1, P<0.01) and a significantly larger trabecular spacing (1.06±0.16 vs. 0.56±0.08 mm, P<0.01). Mean T2* values of osteoporotic patients with and without vertebral fracture showed no significant difference (13.5±3.4 vs. 15.6±3.5 ms, P=0.40). When comparing the mean T2* of the fractured vertebrae, no significant difference could be detected between low-energy osteoporotic fractures and high-energy traumatic fractures (12.6±5.4 vs. 8.1±2.4 ms, P=0.10).

CONCLUSIONS

T2* mapping of vertebral bone marrow using using chemical shift encoding-based water-fat separation allows for assessing osteoporosis as well as the trabecular microstructure and enables a radiation-free differentiation between patients with low-energy osteoporotic and high-energy traumatic vertebral fractures, suggesting its potential as a biomarker for bone fragility.

Sparse-sampling computed tomography for detection of endoleak after endovascular aortic repair (EVAR)

OBJECTIVES

To evaluate sparse sampling computed tomography (SpSCT) for detection of endoleak after endovascular aortic repair (EVAR) at different dose levels in terms of subjective image criteria and diagnostic accuracy.

METHODS

Twenty clinically indicated computed tomography aortic angiography (CTA) scans were used to obtain simulated low-dose scans with 100%, 50%, 25%, 12.5% and 6.25% of the applicated clinical dose, resulting in five dose levels (DL). From full sampling (FS) data sets, every second (2-SpSCT) or fourth (4-SpSCT) projection was used to generate simulated sparse sampling scans. All examinations were evaluated by four blinded radiologists regarding subjective image criteria and diagnostic performance.

RESULTS

Sensitivity was higher than 93% in 4-SpSCT at the 25% DL which is the same as with FS at full dose (100% DL). High accuracies and relative high AUC-values were obtained for 2- and 4-SpSCT down to the 12.5% DL, while for FS similar values were shown down to 25% DL only. Subjective image quality was significantly higher for 4-SpSCT compared to FS at each dose level. More than 90% of all cases were rated with a high or medium confidence for FS and 2-SpSCT at the 50% DL and for 4-SpSCT at the 25% DL. At DL 25% and 12.5%, more cases showed a high confidence using 2- and 4-SpSCT compared with FS.

CONCLUSIONS

Via SpSCT, a dose reduction down to a 25% dose level (mean effective dose of 1.49 mSv in the current study) for CTA is possible while maintaining high image quality and full diagnostic confidence.

Prognostic Assessment in High-Grade Soft-Tissue Sarcoma Patients: A Comparison of Semantic Image Analysis and Radiomics

BACKGROUND

In patients with soft-tissue sarcomas of the extremities, the treatment decision is currently regularly based on tumor grading and size. The imaging-based analysis may pose an alternative way to stratify patients' risk. In this work, we compared the value of MRI-based radiomics with expert-derived semantic imaging features for the prediction of overall survival (OS).

METHODS

Fat-saturated T2-weighted sequences (T2FS) and contrast-enhanced T1-weighted fat-saturated (T1FSGd) sequences were collected from two independent retrospective cohorts (training: 108 patients; testing: 71 patients). After preprocessing, 105 radiomic features were extracted. Semantic imaging features were determined by three independent radiologists. Three machine learning techniques (elastic net regression (ENR), least absolute shrinkage and selection operator, and random survival forest) were compared to predict OS.

RESULTS

ENR models achieved the best predictive performance. Histologies and clinical staging differed significantly between both cohorts. The semantic prognostic model achieved a predictive performance with a C-index of 0.58 within the test set. This was worse compared to a clinical staging system (C-index: 0.61) and the radiomic models (C-indices: T1FSGd: 0.64, T2FS: 0.63). Both radiomic models achieved significant patient stratification.

CONCLUSIONS

T2FS and T1FSGd-based radiomic models outperformed semantic imaging features for prognostic assessment.

Opportunistic QCT Bone Mineral Density Measurements Predicting Osteoporotic Fractures: A Use Case in a Prospective Clinical Cohort

PURPOSE

To assess whether volumetric vertebral bone mineral density (BMD) measured with opportunistic quantitative computed tomography (QCT) (i.e., CT acquired for other reasons) can predict osteoporotic fracture occurrence in a prospective clinical cohort and how this performs in comparison to dual-energy X-ray absorptiometry (DXA) measurements.

METHODS

In the database of our fracture liaison service, 58 patients (73 ± 11 years, 72% women) were identified that had at least one prevalent low-energy fracture and had undergone CT of the spine. BMD was determined by converting HU using scanner-specific conversion equations. Baseline DXA was available for 31 patients. During a 3-year follow-up, new fractures were diagnosed either by (i) recent in-house imaging or (ii) clinical follow-up with validated external reports. Associations were assessed using logistic regression models, and cut-off values were determined with ROC/Youden analyses.

RESULTS

Within 3 years, 20 of 58 patients presented new low-energy fractures (34%). Mean QCT BMD of patients with fractures was significantly lower (56 ± 20 vs. 91 ± 38 mg/cm³; p = 0.003) and age was higher (77 ± 10 vs. 71 ± 11 years; p = 0.037). QCT BMD was significantly associated with the occurrence of new fractures, and the OR for developing a new fracture during follow-up was 1.034 (95% CI, 1.010-1.058, p = 0.005), suggesting 3% higher odds for every unit of BMD decrease (1 mg/cm³). Age and sex showed no association. For the differentiation between patients with and without new fractures, ROC showed an AUC of 0.76 and a Youden's Index of J = 0.48, suggesting an optimal cut-off value of 82 mg/cm³. DXA T-scores showed no significant association with fracture occurrence in analogous regression models.

CONCLUSION

In this use case, opportunistic BMD measurements attained through QCT predicted fractures during a 3-year follow-up. This suggests that opportunistic measurements are useful to reduce the diagnostic gap and evaluate the fracture risk in osteoporotic patients.

Lah is a transmembrane protein and requires Spa10 for stable positioning of Woronin bodies at the septal pore of Aspergillus fumigatus

Woronin bodies are specialized, fungal-specific organelles that enable an immediate closure of septal pores after injury to protect hyphae from excessive cytoplasmic bleeding. In most Ascomycetes, Woronin bodies are tethered at the septal pore by so-called Lah proteins. Using the pathogenic mold Aspergillus fumigatus as a model organism, we show that the C-terminal 288 amino acids of Lah (LahC₂₈₈) bind to the rim of the septal pore. LahC₂₈₈ essentially consists of a membrane spanning region and a putative extracellular domain, which are both required for the targeting to the septum. In an A. fumigatus rho4 deletion mutant that has a severe defect in septum formation, LahC₂₈₈ is recruited to spot-like structures in or at the lateral membrane. This suggests that LahC is recruited before Rho4 starts to govern the septation process. Accordingly, we found that in wild type hyphae Lah is bound before a cross-wall emerges and thus enables a tethering of Woronin bodies at the site of the newly formed septum. Finally, we identified Spa10, a member of a recently described family of septal pore-associated proteins, as a first protein that directly or indirectly interacts with LahC to allow a stable positioning of Woronin bodies at the mature septum.

Functional characterization of the Woronin body protein WscA of the pathogenic mold Aspergillus fumigatus

Woronin bodies are fungal-specific organelles that seal damaged hyphal compartments and thereby contribute to the stress resistance and virulence of filamentous fungi. In this study, we have characterized the Aspergillus fumigatus Woronin body protein WscA. WscA is homologous to Neurospora crassa WSC, a protein that was shown to be important for biogenesis, segregation and positioning of Woronin bodies. WscA and WSC both belong to the Mpv17/PMP22 family of peroxisomal membrane proteins. An A. fumigatus ΔwscA mutant is unable to form Woronin bodies, and HexA, the protein that forms the crystal-like core of Woronin bodies, accumulates in large peroxisomes instead. The ΔwscA mutant showed no defect in segregation of HexA containing organelles, as has been reported for the corresponding N. crassa mutant. In the peroxisomes of the A. fumigatus mutant, HexA assembles into compact, donut-shaped structures. Experiments with GFP fusion proteins revealed that WscA function is highly sensitive to these modifications, in particular to an N-terminal fusion of GFP. In N. crassa, WSC was shown to be essentially required for Woronin body positioning, but the respective domain is not conserved in most other Pezizomycotina, including A. fumigatus. We have recently found evidence that HexA may have a direct role in WB positioning, since a HexA-GFP fusion protein, lacking a functional PTS1 motif, is efficiently recruited to the septal pore. In the current study we show that this targeting of HexA-GFP is independent of WscA.