Automated segmentation of trabecular and cortical bone from proton density weighted MRI of the knee

Can magnetic resonance imaging accurately and reliably measure humeral cortical thickness?

Background

Historically, imaging osseous detail in three dimensions required a computed tomography (CT) scan with ionizing radiation that poorly visualizes the soft tissues. The purpose of this study was to determine the accuracy and reliability of ultrashort echo time (UTE) magnetic resonance imaging (MRI) in measuring humeral cortical thickness and cancellous density as compared with CT.

Methods

This was a comparative radiographic study in nine cadavers, each of which underwent CT and UTE MRI. On images aligned to the center of the humeral shaft, anterior, posterior, medial, and lateral humeral cortical thickness was measured 5, 10, and 15 cm distal to the top of the head. Cancellous density was measured as signal within a 1-cm diameter region of interest in the center of the head, the subtuberosity head, the subarticular head, and the subarticular glenoid vault. Glenoid cortical thickness was measured at the center of the glenoid. Cortical measurements were compared using mean differences and 95% confidence intervals, paired Student’s t-tests, and intraclass correlation coefficients (ICCs). We compared cancellous measurements using Pearson’s correlation coefficients. For all measurements, we calculated interobserver and intraobserver reliability using ICCs with 0.75 as the lower limit for acceptability.

Results

With regard to accuracy, for humeral cortical thickness measurements, there were no significant differences between MRI and CT measures, and ICCs were >0.75. The glenoid cortical thickness ICC was <0.75. There was no significant correlation between the cancellous signal on MRI and on CT in any region. For both MRI and CT, interobserver reliability and intraobserver reliability were acceptable (ie, >0.75) for almost all humeral cortical thickness measures.

Conclusion

UTE MRI can reliably and accurately measure humeral cortical thickness, but cannot accurately measure cancellous density or accurately and reliably measure glenoid cortical thickness.

Automated cortical thickness measurement of the mandibular condyle head on CBCT images using a deep learning method

This study proposes a deep learning model for cortical bone segmentation in the mandibular condyle head using cone-beam computed tomography (CBCT) and an automated method for measuring cortical thickness with a color display based on the segmentation results. In total, 12,800 CBCT images from 25 normal subjects, manually labeled by an oral radiologist, served as the gold-standard. The segmentation model combined a modified U-Net and a convolutional neural network for target region classification. Model performance was evaluated using intersection over union (IoU) and the Hausdorff distance in comparison with the gold standard. The second automated model measured the cortical thickness based on a three-dimensional (3D) model rendered from the segmentation results and presented a color visualization of the measurements. The IoU and Hausdorff distance showed high accuracy (0.870 and 0.928 for marrow bone and 0.734 and 1.247 for cortical bone, respectively). A visual comparison of the 3D color maps showed a similar trend to the gold standard. This algorithm for automatic segmentation of the mandibular condyle head and visualization of the measured cortical thickness as a 3D-rendered model with a color map may contribute to the automated quantification of bone thickness changes of the temporomandibular joint complex on CBCT.

Feasibility of using a single MRI acquisition for fiducial marker localization and synthetic CT generation towards MRI-only prostate radiation therapy treatment planning

Purpose.To investigate the feasibility of using a single MRI acquisition for fiducial marker identification and synthetic CT (sCT) generation towards MRI-only treatment planning for prostate external beam radiation therapy (EBRT).Methods.Seven prostate cancer patients undergoing EBRT, each with three implanted gold fiducial markers, participated in this study. In addition to the planning CT scan, all patients were scanned on a 3 T MR scanner with a 3D double-echo gradient echo (GRE) sequence. Quantitative susceptibility mapping (QSM) was performed for marker localization. QSM-derived marker positions were compared to those from CT. The bulk density assignment technique for sCT generation was adopted. The magnitude GRE images were segmented into muscle, bone, fat, and air using a combination of unsupervised intensity-based classification of soft tissue and convolutional neural networks (CNN) for bone segmentation.Results.All implanted markers were visualized and accurately identified (average error: 0.7 ± 0.5 mm). QSM generated distinctive contrast for hemorrhage, calcifications, and gold fiducial markers. The estimated susceptibility/HU values on QSM/CT for gold and calcifications were 31.5 ± 2.9 ppm/1220 ± 100 HU and 14.6 ± 0.9 ppm/440 ± 100 HU, respectively. The intensity-based soft tissue classification resulted in an average Dice score of 0.97 ± 0.02; bone segmentation using CNN resulted in an average Dice score of 0.93 ± 0.03.Conclusion.This work indicates the feasibility of simultaneous fiducial marker identification and sCT generation using a single MRI acquisition. Future works includes evaluation of the proposed method in a large cohort of patients with optimized acquisition parameters as well as dosimetric evaluations.

Novel Autosegmentation Spatial Similarity Metrics Capture the Time Required to Correct Segmentations Better Than Traditional Metrics in a Thoracic Cavity Segmentation Workflow

Automated segmentation templates can save clinicians time compared to de novo segmentation but may still take substantial time to review and correct. It has not been thoroughly investigated which automated segmentation-corrected segmentation similarity metrics best predict clinician correction time. Bilateral thoracic cavity volumes in 329 CT scans were segmented by a UNet-inspired deep learning segmentation tool and subsequently corrected by a fourth-year medical student. Eight spatial similarity metrics were calculated between the automated and corrected segmentations and associated with correction times using Spearman’s rank correlation coefficients. Nine clinical variables were also associated with metrics and correction times using Spearman’s rank correlation coefficients or Mann–Whitney U tests. The added path length, false negative path length, and surface Dice similarity coefficient correlated better with correction time than traditional metrics, including the popular volumetric Dice similarity coefficient (respectively ρ = 0.69, ρ = 0.65, ρ =  − 0.48 versus ρ =  − 0.25; correlation p values < 0.001). Clinical variables poorly represented in the autosegmentation tool’s training data were often associated with decreased accuracy but not necessarily with prolonged correction time. Metrics used to develop and evaluate autosegmentation tools should correlate with clinical time saved. To our knowledge, this is only the second investigation of which metrics correlate with time saved. Validation of our findings is indicated in other anatomic sites and clinical workflows. Novel spatial similarity metrics may be preferable to traditional metrics for developing and evaluating autosegmentation tools that are intended to save clinicians time.