Hand bone segmentation in radioabsorptiometry images for computerised bone mass assessment

Convolutional Neural Network-Based Humerus Segmentation and Application to Bone Mineral Density Estimation from Chest X-ray Images of Critical Infants

Measuring bone mineral density (BMD) is important for surveying osteopenia in premature infants. However, the clinical availability of dual-energy X-ray absorptiometry (DEXA) for standard BMD measurement is very limited, and it is not a practical technique for critically premature infants. Developing alternative approaches for DEXA might improve clinical care for bone health. This study aimed to measure the BMD of premature infants via routine chest X-rays in the intensive care unit. A convolutional neural network (CNN) for humeral segmentation and quantification of BMD with calibration phantoms (QRM-DEXA) and soft tissue correction were developed. There were 210 X-rays of premature infants evaluated by this system, with an average Dice similarity coefficient value of 97.81% for humeral segmentation. The estimated humerus BMDs (g/cm³; mean ± standard) were 0.32 ± 0.06, 0.37 ± 0.06, and 0.32 ± 0.09, respectively, for the upper, middle, and bottom parts of the left humerus for the enrolled infants. To our knowledge, this is the first pilot study to apply a CNN model to humerus segmentation and to measure BMD in preterm infants. These preliminary results may accelerate the progress of BMD research in critical medicine and assist with nutritional care in premature infants.

Dual-Energy X-Ray Absorptiometry for Measurement of Phalangeal Bone Mineral Density on a Slot-Scanning Digital Radiography System

OBJECTIVE

In this paper, we assess the feasibility of using two detectors in a slot-scanning digital radiography system to acquire images for measuring bone mineral density (BMD) of the middle phalanx of the middle finger using dual-energy X-ray absorptiometry (DXA).

METHODS

Simulations were used to evaluate the spectral separation of the low- and high-energy spectra and detective quantum efficiency was used for assessing image quality. Scan parameters were chosen to optimize spectral separation, image quality, and radiation dose. We introduce the measurement of volumetric BMD (vBMD) using basis material decomposition. We assess the accuracy of our methods by comparing measurements taken using bone images against reference data derived from subsequent incineration of the bones. In vivo scans were conducted to evaluate the system precision (repeatability) and agreement with a clinical densitometer.

RESULTS

Average errors for bone mineral content (BMC), areal BMD (aBMD), and vBMD were 4.85%, 5.49%, and 12.77%, respectively. Our system had good agreement with a clinical densitometer based on concordance correlation coefficient values of 0.92 and 0.98 for aBMD and BMC, respectively. Precision studies yielded coefficient of variation (CV) values of 1.35% for aBMD, 1.48% for BMC, and 1.80% for vBMD. The CV values of all measurements were within 2%, indicating that the methods have clinically acceptable precision.

CONCLUSION

We conclude that our techniques yield bone measurements with high accuracy, clinically acceptable precision, and good agreement with a clinical densitometer.

SIGNIFICANCE

We have shown the clinical potential of phalangeal DXA measurements of aBMD and vBMD on a slot-scanning digital radiography system.

Computer-based measurements of joint space analysis using metacarpal morphometry in hand radiograph for evaluation of rheumatoid arthritis

AIM AND OBJECTIVES

The aim and objectives are as follows: (i) to perform an automated segmentation of the hand from radiographs using a dual tree complex wavelet-based watershed algorithm; ii) to compare the measured statistical features of the joint space of the hand using gray level co-occurrence matrix (GLCM) method with standard diagnostic parameters of rheumatoid arthritis (RA).

METHODS

Fifty-three patients with RA and 17 age- and sex-matched healthy controls were included in the study. The erythrocyte sedimentation rate (ESR), C-reactive protein, rheumatoid factor, health assessment questionnaire score (HAQ), disease activity score (DAS) and hand radiographs of all the subjects were obtained. Joint space width and cortical thickness were measured in metacarpophalangeal joints (MCP) and metacarpal bone semi-automatically using MIMICS software. Dual tree complex wavelet transform-based watershed algorithm was applied for automated segmentation, and feature extraction was performed using the GLCM method in hand radiographs of the total population.

RESULTS

In the RA group (n = 53), the joint space width measured in the MCP1, MCP3, MCP5 of the hand were reduced significantly (P < 0.01) by 16.4%, 15.6%, and 17.5%, respectively compared to the normal group (n = 17). The measured combined cortical thickness at the second, third and fourth metacarpal bones of the hand were reduced significantly (P < 0.01) by 9.5%, 12% and 8%, respectively in the RA group compared to the normal group.

CONCLUSION

The dual tree complex wavelet transform-based watershed algorithm provided effective segmentation in the digitized hand radiographs. The standard diagnostic parameters for RA were highly correlated with measured statistical features at MCP3 hand joint in the total population studied.

Computed digital absorptiometry for measurement of phalangeal bone mineral mass on a slot-scanning digital radiography system

Computed digital absorptiometry is a low-cost and low-radiation technique for rapid measurement of phalangeal bone mineral mass. We implement and evaluate this technique on a slot-scanning radiography system. Results, based on measurements of excised phalangeal bones, indicate that the technique has potential for use in clinical assessment of osteoporosis.

INTRODUCTION

The current gold standard method for bone assessment in the diagnosis of osteoporosis requires specialised and expensive machines, highly trained personnel to conduct the examination and is available only at specialist centres. The technique, termed dual-energy X-ray absorptiometry (DXA), involves taking a bone mineral density measurement at the femur or lumbar spine. Measurements of bone at peripheral sites such as the phalanges using DXA and other techniques have been shown to have potential use in the diagnosis of osteoporosis. Computed digital absorptiometry (CDA) is a low-cost, low-radiation radiographic technique for assessing phalangeal bone mineral mass. It uses an aluminium step wedge as a calibration device to compute bone mineral mass in units of equivalent aluminium thickness. In this study, we assess the feasibility of using CDA on a slot-scanning radiography system for measuring phalangeal bone mineral mass.

METHODS

We implement and evaluate fully automated computed digital absorptiometry (CDA) of the middle phalanx of the middle finger on a slot-scanning radiography system.

RESULTS

The ash weight of incinerated bones was measured and shown to have a correlation of 0.92 with CDA-derived bone mineral mass. CDA measurements had a coefficient of variation of 0.26%, indicating high precision.

CONCLUSION

We conclude, based on these results, that CDA on a slot-scanning radiography machine may be useful for clinical assessment of osteoporosis.

Statistical shape and appearance models in osteoporosis

Statistical models (SMs) of shape (SSM) and appearance (SAM) have been acquiring popularity in medical image analysis since they were introduced in the early 1990s. They have been primarily used for segmentation, but they are also a powerful tool for 3D reconstruction and classification. All these tasks may be required in the osteoporosis domain, where fracture detection and risk estimation are key to reducing the mortality and/or morbidity of this bone disease. In this article, we review the different applications of SSMs and SAMs in the context of osteoporosis, and it concludes with a discussion of their advantages and disadvantages for this application.

Automated bone age assessment: motivation, taxonomies, and challenges

Bone age assessment (BAA) of unknown people is one of the most important topics in clinical procedure for evaluation of biological maturity of children. BAA is performed usually by comparing an X-ray of left hand wrist with an atlas of known sample bones. Recently, BAA has gained remarkable ground from academia and medicine. Manual methods of BAA are time-consuming and prone to observer variability. This is a motivation for developing automated methods of BAA. However, there is considerable research on the automated assessment, much of which are still in the experimental stage. This survey provides taxonomy of automated BAA approaches and discusses the challenges. Finally, we present suggestions for future research.

Model-based segmentation of the middle phalanx in digital radiographic images of the hand

We present techniques for segmenting the middle phalanx of the middle finger in digital radiographic images using deformable models and active shape models (ASMs). The result of segmentation may be used in the estimation of bone mineral density which in turn may be used in the diagnosis of osteoporosis. A technique for minimizing user dependence is described. The segmentation accuracy of the two methods is assessed by comparing contours produced by the algorithms to those produced by manual segmentation, using the Hausdorff distance measure. The ASM technique produces more accurate segmentation.

An artifacts removal post-processing for epiphyseal region-of-interest (EROI) localization in automated bone age assessment (BAA)

Background

Segmentation is the most crucial part in the computer-aided bone age assessment. A well-known type of segmentation performed in the system is adaptive segmentation. While providing better result than global thresholding method, the adaptive segmentation produces a lot of unwanted noise that could affect the latter process of epiphysis extraction.

Methods

A proposed method with anisotropic diffusion as pre-processing and a novel Bounded Area Elimination (BAE) post-processing algorithm to improve the algorithm of ossification site localization technique are designed with the intent of improving the adaptive segmentation result and the region-of interest (ROI) localization accuracy.

Results

The results are then evaluated by quantitative analysis and qualitative analysis using texture feature evaluation. The result indicates that the image homogeneity after anisotropic diffusion has improved averagely on each age group for 17.59%. Results of experiments showed that the smoothness has been improved averagely 35% after BAE algorithm and the improvement of ROI localization has improved for averagely 8.19%. The MSSIM has improved averagely 10.49% after performing the BAE algorithm on the adaptive segmented hand radiograph.

Conclusions

The result indicated that hand radiographs which have undergone anisotropic diffusion have greatly reduced the noise in the segmented image and the result as well indicated that the BAE algorithm proposed is capable of removing the artifacts generated in adaptive segmentation.

Model-based segmentation of flexor tendons from magnetic resonance images of finger joints

Trigger finger is a common hand disease, causing swelling, painful popping and clicking in moving the affected finger joint. To better evaluate patients with trigger finger, segmentation of flexor tendons from magnetic resonance (MR) images of finger joints, which can offer detailed structural information of tendons to clinicians, is essential. This paper presents a novel model-based method with three stages for automatically segmenting the flexor tendons. In the first stage, a set of tendon contour models (TCMs) is initialized from the most proximal cross-sectional image via two-step ellipse estimation. Each of the TCMs is then propagated to its distally adjacent image by affine registration. The propagation is sequentially performed along the proximal-distal direction until the most distal image is reached, as the second stage of segmentation. The TCMs on each cross-sectional image are refined in the last stage with the snake deformation. MR volumes of three subjects were used to validate the segmentation accuracy. Compared with the manual results, our method showed good accuracy with small average margins of errors (within 0.5 mm) and large overlapping ratio (dice similarity coefficient above 0.8). Overall, the proposed method has great potential for morphological change assessment of flexor tendons and pulley-tendon system modeling for image guided surgery.

Automated segmentation for patella from lateral knee X-ray images

X-ray image segmentation is an important issue in medical image analysis. Due to inconsistent X-ray absorption, the intensities are usually unevenly distributed and noisy in the processed organ, thus the object segmentation becomes difficult. In this paper we propose a new segmentation method for patella from the lateral knee X-ray images based on the active shape model (ASM). At first, a patella shape model is constructed by principal component analysis (PCA) of corresponding landmarks obtained from a set of training shape. As the knee X-ray image usually contains many anatomical structures, we design a strategy based on edge tracing to place the initial shape model as close to the patella boundary as possible. Then, the shape model is deformed and fitted to the patella boundary by using a dual-optimization approach that includes a genetic algorithm (GA) to get the global geometric transform and ASM to deform the shape model iteratively. Consequently, the proposed method can cope with different knee X-ray images and can segment the patella in an automatic procedure. In the experiment, 20 images were tested and promising results are obtained by the proposed method. This method is found useful for the clinical evaluation and biomechanical study of knee.

Learning active shape models for bifurcating contours

Statistical shape models are often learned from examples based on landmark correspondences between annotated examples. A method is proposed for learning such models from contours with inconsistent bifurcations and loops. Automatic segmentation of tibial and femoral contours in knee X-ray images is investigated as a step towards reliable, quantitative radiographic analysis of osteoarthritis for diagnosis and assessment of progression. Results are presented using various features, the Mahalanobis distance, distance weighted K-nearest neighbours, and two relevance vector machine-based methods as quality of fit measure.