Chi-Wen Hsieh

Long-term secular trend of skeletal maturation of Taiwanese children between agricultural (1960s) and contemporary (after 2000s) generations using the Tanner-Whitehouse 3 (TW3) method.

Abstract Background: The Tanner-Whitehouse (TW) method is one of the well-known techniques in determining the bone age. Objective: According to the objectivity of TW3, the secular trend was investigated to discover whether the skeletal maturation of Taiwanese children between two generations was different. Subjects and methods: The large-scale database of Taiwan was collected. The first group, called mid-1960s, included 265 boys and 295 girls in the agricultural generation (between 1966 and 1967). The second group, called mid-2000s, includes 114 boys and 616 girls in the contemporary generation (after 2000s). The bone age was determined by three radiologists using the carpals-only system of the TW3 method and by two physicians using the Greulich and Pyle method. A comparison of the means (independent-samples t-test) was applied by examining the difference of the children’s skeletal maturation between the two generations in the same chronological age. The significant difference was considered while the p-value was 0.05 or less (95% confidence interval). Results: A significant difference of the mean bone age (by, on average, three radiologists using the TW3 method) between the mid-1960s and mid-2000s in the same gender and chronological age was presented by the independent-samples t-test (p<0.001 with 95% confidence interval), and the bone age, determined by the TW3 method, of the mid-2000s group was higher than that of the mid-1960s group. This scenario corresponded with the children’s bone age determined by pediatricians. Besides, it deserved to notice that the bone age of boys in the mid-2000s was larger than that of the girls in the mid-1960s. Furthermore, by comparing the environmental condition, we suspect that the difference of bone age of children between the two generations was attributed to the discrepancy in nutrition and socioeconomic variation during the four decades in Taiwan. Conclusion: The study presents that the secular trend of skeletal maturation of children in the mid-2000s is faster than that in the mid-1960s.

Simplified radius, ulna, and short bone-age assessment procedure using grouped-Tanner-Whitehouse method

Background:  The Tanner–Whitehouse III (TW3) method is popular for assessing childrens bone age, but it is time‐consuming in clinical settings; to simplify this, a grouped‐TW algorithm (GTA) was developed.

Fast and fully automatic phalanx segmentation using a grayscale-histogram morphology algorithm

Bone age assessment is a common radiological examination used in pediatrics to diagnose the discrepancy between the skeletal and chronological age of a child; therefore, it is beneficial to develop a computer-based bone age assessment to help junior pediatricians estimate bone age easily. Unfortunately, the phalanx on radiograms is not easily separated from the background and soft tissue. Therefore, we proposed a new method, called the grayscale-histogram morphology algorithm, to segment the phalanges fast and precisely. The algorithm includes three parts: a tri-stage sieve algorithm used to eliminate the background of hand radiograms, a centroid-edge dual scanning algorithm to frame the phalanx region, and finally a segmentation algorithm based on disk traverse-subtraction filter to segment the phalanx. Moreover, two more segmentation methods: adaptive two-mean and adaptive two-mean clustering were performed, and their results were compared with the segmentation algorithm based on disk traverse-subtraction filter using five indices comprising misclassification error, relative foreground area error, modified Hausdorff distances, edge mismatch, and region nonuniformity. In addition, the CPU time of the three segmentation methods was discussed. The result showed that our method had a better performance than the other two methods. Furthermore, satisfactory segmentation results were obtained with a low standard error.

A novel image enhancement approach for Phalanx and Epiphyseal/metaphyseal segmentation based on hand radiographs

Precise and effective bony segmentation from soft tissues on hand radiograph is very critical in the bone age assessment, especially for separating phalanx and epiphyseal/metaphyseal region of interests (EMROIs) from the background and soft tissue. Our aim is to locate and segment the EMROIs based on childrens hand radiographs, in which the images are with nonuniform background. Furthermore, a great destructive noise may suppress the segmentation process. The proposed approach combines a series of image preprocessing procedures to crop the left-hand radiograph and EMROIs. The experiments were conducted on different parts of knuckles, such as distal, middle, and proximal. Finally, the segmentation results demonstrate that the proposed schemes can segment the EMROIs automatically and effectively for the extraction of characteristics of bone age assessment.

A study of video-based droplet detection

The disdrometer system is a new imaged-based technique developed to measure the shape of droplets. This system contains three components: a CCD camera for capturing video images, a backlight source for illuminating drops and enhancing images, and a control unit for data acquisition and function control. In our preliminary study, we replaced raindrops with a free fall of droplets from a homemade liquid dropper. Subsequently, the high-speed CCD camera captured images of the falling droplets. We controlled the exposure time to 1 ms, which resulted in the falling droplets blurring in the assessment. To detect the droplet contours precisely, we used Pascals theorem to determine the best-fit ellipse shape. We concluded that the proposed image-based disdrometer system provides a promising approach to measure and monitor the falling droplets. Moreover, this instrument can be less expensive as it only requires basic maintenance for long-lasting daily use.

Bone Age Pre-estimation Using Partial Least Squares Regression Analysis with a Priori Knowledge

Computerized bone age estimation was widely developed during near decade. There are many techniques were applied to this filed. In the study, a new concept, the partial least squares (PLS) regression analysis was proposed to pre-estimate the bone age as a priori knowledge. PLS is a multivariate statistical analysis method and can find combinations of the predictors that have a large covariance with the response values. One hundred and eleven radiograms with the third finger of the left hand were retrieved from DICOM system of Department of Radiology, Taipei Veterans General Hospital. The projection data of radiograms with the third finger were calculated, and then the normalized process was implemented. The discrete cosine transform (DCT) was employed to obtain components from projection data as features in this study. The first 30 components of DCT were selected as features. PLS was utilized to separate these features into four stages of skeleton maturation of children as a new application. The result of PLS regression presented that the average residual is 0.5. The study demonstrated a significant result that the PLS can successfully pre-estimate the bone age as a priori knowledge.

An easy-to-use TW3 system for assessing bone age

Precise and effective bone age assessment is important for pediatricians to diagnose children with endocrinological problems and growth disorders. The Tanner and Whitehouse III (TW3) method for determining bone age involves detailed shape analyses of several bones of interest and assigning corresponding scores, but the complexity and time-consuming nature make it hard to apply in clinic. Therefore, our project is to develop an easy-to-use graphical user interface which can automatically localize and segment the phalanx region of interests (PROIs) and the epiphyseal/metaphyseal region of interests (EMROIs) of the left-hand radiogram, and then a TW3 scoring system is provided to the user to make a decision. Based on this system, a set of image processing procedures was proposed for the segmentation of PROIs and EMROIs, and then 8 EMORIs were consecutively popped up for the evaluation of TW3 scores. Subsequently, each score of 8 EMROIs was accumulated automatically until the operations were completed. We believed that the easy-to-use TW3 system serves a better operational tool for the pediatricians.

An adaptive level set method for improving image segmentation

Segmentation is one of the most significant techniques in image processing studies. But the major limitations of low contrast, ambiguous boundary and complex morphology of images usually lead to unsatisfactory segmentation, especially in heavily noisy medical images. To address this problem, an attempt by improving Distance Regularized Level Set Evolution (DRLSE) model has been made to build up an effective and accurate active contour model to improve the segmentation ability. In this study, a balance concept between the pushing force and image’s contour strength of DRLSE model was adapted for contour segmentations, so the contour convergent/divergent force will be re-defined to make it more feasible for complicated boundaries. The proposed concept integrates the bilateral filtering and Canny contour, and the improved edge indicator function into the DRLSE model. The most important step is to adapt the improved edge indicator function appropriately with the Canny contour. To verify our proposed DRLSE scheme, different testing images are used. The experimental results show that the proposed approach has been examined and tested successfully in most cases. Consequently, the developed algorithm including the DRLSE model combing the Canny contour with an improved g edge indicator, demonstrate the effectiveness and reliability, especially in the testing samples with weak edges and complex topologies.

Automatic inspection apparatus based on high-speed image analysis: a new characterization technique for particle flow analysis

To better understand the mechanism of ejecting spray droplet and raindrop, a high-speed image analysis for droplet characterization technique was proposed in this study. In our design, three basic units containing a high-speed CCD camera, a light source, and automatic image acquisition and data processing unit, are integrated to assess the droplet motion. Next, a set of multistage image processing algorithms, the droplet detection and droplet tracking, were developed and conducted to analyze the droplet trajectories of the grabbed frames. By controlling a very short exposure time, the droplet size and velocity can be obtained to generate the droplet size distribution (DSD). To examine our proposed approaches, a commercially available spray nozzle and raindrop are used to evaluate the effectiveness. Meanwhile, the pair-matching rate is also involved to validate the recognition rate of the droplet tracking algorithm. Using this proposed framework, it can be used to provide valuable data to assess the spray droplet or the precipitation status by their droplet size to droplet velocity.

Differentiating malignant from benign breast tumors on acoustic radiation force impulse imaging using fuzzy-based neural networks with principle component analysis

Many modalities have been developed as screening tools for breast cancer. A new screening method called acoustic radiation force impulse (ARFI) imaging was created for distinguishing breast lesions based on localized tissue displacement. This displacement was quantitated by virtual touch tissue imaging (VTI). However, VTIs sometimes express reverse results to intensity information in clinical observation. In the study, a fuzzy-based neural network with principle component analysis (PCA) was proposed to differentiate texture patterns of malignant breast from benign tumors. Eighty VTIs were randomly retrospected. Thirty four patients were determined as BI-RADS category 2 or 3, and the rest of them were determined as BI-RADS category 4 or 5 by two leading radiologists. Morphological method and Boolean algebra were performed as the image preprocessing to acquire region of interests (ROIs) on VTIs. Twenty four quantitative parameters deriving from first-order statistics (FOS), fractal dimension and gray level co-occurrence matrix (GLCM) were utilized to analyze the texture pattern of breast tumors on VTIs. PCA was employed to reduce the dimension of features. Fuzzy-based neural network as a classifier to differentiate malignant from benign breast tumors. Independent samples test was used to examine the significance of the difference between benign and malignant breast tumors. The area Az under the receiver operator characteristic (ROC) curve, sensitivity, specificity and accuracy were calculated to evaluate the performance of the system. Most all of texture parameters present significant difference between malignant and benign tumors with p-value of less than 0.05 except the average of fractal dimension. For all features classified by fuzzy-based neural network, the sensitivity, specificity, accuracy and Az were 95.7%, 97.1%, 95% and 0.964, respectively. However, the sensitivity, specificity, accuracy and Az can be increased to 100%, 97.1%, 98.8% and 0.985, respectively if PCA was performed to reduce the dimension of features. Patterns of breast tumors on VTIs can effectively be recognized by quantitative texture parameters, and differentiated malignant from benign lesions by fuzzy-based neural network with PCA.