Automatic segmentation of hepatocellular carcinoma on dynamic contrast-enhanced MRI based on deep learning

Objective. Precise hepatocellular carcinoma (HCC) detection is crucial for clinical management. While studies focus on computed tomography-based automatic algorithms, there is a rareness of research on automatic detection based on dynamic contrast enhanced (DCE) magnetic resonance imaging. This study is to develop an automatic detection and segmentation deep learning model for HCC using DCE.Approach: DCE images acquired from 2016 to 2021 were retrospectively collected. Then, 382 patients (301 male; 81 female) with 466 lesions pathologically confirmed were included and divided into an 80% training-validation set and a 20% independent test set. For external validation, 51 patients (42 male; 9 female) in another hospital from 2018 to 2021 were included. The U-net architecture was modified to accommodate multi-phasic DCE input. The model was trained with the training-validation set using five-fold cross-validation, and furtherly evaluated with the independent test set using comprehensive metrics for segmentation and detection performance. The proposed automatic segmentation model consisted of five main steps: phase registration, automatic liver region extraction using a pre-trained model, automatic HCC lesion segmentation using the multi-phasic deep learning model, ensemble of five-fold predictions, and post-processing using connected component analysis to enhance the performance to refine predictions and eliminate false positives.Main results. The proposed model achieved a mean dice similarity coefficient (DSC) of 0.81 ± 0.11, a sensitivity of 94.41 ± 15.50%, a precision of 94.19 ± 17.32%, and 0.14 ± 0.48 false positive lesions per patient in the independent test set. The model detected 88% (80/91) HCC lesions in the condition of DSC > 0.5, and the DSC per tumor was 0.80 ± 0.13. In the external set, the model detected 92% (58/62) lesions with 0.12 ± 0.33 false positives per patient, and the DSC per tumor was 0.75 ± 0.10.Significance.This study developed an automatic detection and segmentation deep learning model for HCC using DCE, which yielded promising post-processed results in accurately identifying and delineating HCC lesions.

Mean S1 inlet and outlet view angles are not safe for all individuals according to three-dimensional tomographic measurements

Although there are many studies evaluating optimal inlet and outlet angles required for the correct placement of S1 iliosacral screws, there is no study evaluating reliability and feasibility of these angles for all individuals on three-dimensional (3D) anatomical models. A total of 100 women and 100 men were selected randomly. A vertical line was created according to long axis of the tomography device on which patient was lying in supine position. The automatized best-fit planes were created on superior and inferior endplates, anterior cortex including notch region and posterior cortex of first sacral vertebrae using 3D imaging software to measure mean inlet and outlet angles. We observed no statistically significant difference between gender groups in terms of inlet and outlet angles. Mean inlet view is obtained for anterior cortex of S1 in 22.5 ± 9.5° and for posterior cortex in 46.5 ± 9.3°. Mean fluoroscopic view angle of S1 for superior outlet is 40.3 ± 7.6 and for inferior outlet is 46.9 ± 8.8. Mean anterior and posterior S1 inlet view angles do not accurately visualize anterior cortex of 74 (37%) and posterior cortex of 66 (33%) individuals. Mean superior and inferior S1 outlet view angles do not accurately visualize superior endplate of 74 (37%) and inferior endplate of 56 (28%) individuals. Due to individual alterations of spatial position of sacrum, mean inlet and outlet view angles of S1 are not sufficient to visualize the iliosacral screws under fluoroscopy in many individuals.

Sixty-four-fold data reduction of chest radiographs using a super-resolution convolutional neural network

OBJECTIVES

To develop and validate a super-resolution (SR) algorithm generating clinically feasible chest radiographs from 64-fold reduced data.

METHODS

An SR convolutional neural network was trained to produce original-resolution images (output) from 64-fold reduced images (input) using 128 × 128 patches (n = 127 030). For validation, 112 radiographs-including those with pneumothorax (n = 17), nodules (n = 20), consolidations (n = 18), and ground-glass opacity (GGO; n = 16)-were collected. Three image sets were prepared: the original images and those reconstructed using SR and conventional linear interpolation (LI) using 64-fold reduced data. The mean-squared error (MSE) was calculated to measure similarity between the reconstructed and original images, and image noise was quantified. Three thoracic radiologists evaluated the quality of each image and decided whether any abnormalities were present.

RESULTS

The SR-images were more similar to the original images than the LI-reconstructed images (MSE: 9269 ± 1015 vs. 9429 ± 1057; P = .02). The SR-images showed lower measured noise and scored better noise level by three radiologists than both original and LI-reconstructed images (Ps < .01). The radiologists' pooled sensitivity with the SR-reconstructed images was not significantly different compared with the original images for detecting pneumothorax (SR vs. original, 90.2% [46/51] vs. 96.1% [49/51]; P = .19), nodule (90.0% [54/60] vs. 85.0% [51/60]; P = .26), consolidation (100% [54/54] vs. 96.3% [52/54]; P = .50), and GGO (91.7% [44/48] vs. 95.8% [46/48]; P = .69).

CONCLUSIONS

SR-reconstructed chest radiographs using 64-fold reduced data showed a lower noise level than the original images, with equivalent sensitivity for detecting major abnormalities.

ADVANCES IN KNOWLEDGE

This is the first study applying super-resolution in data reduction of chest radiographs.

A comparative study of the effectiveness of photogrammetric versus manual anthropometric measurements

BACKGROUND

The accurate measurement of the human body is essential when it comes to designing agricultural tools and equipment that can effectively accommodate and interact with individuals when performing a task. The traditional method for measuring an individual's body measurements is highly complex and requires two or more skilled individuals and reliable measurement tools. Finding a new approach that is speedier, more precise, and less expensive than current methods is therefore necessary.

OBJECTIVE

This study aims to develop an inexpensive novel photogrammetric anthropometric measurement setup that can extract the dimensions of an individual subject irrespective of their body shape.

METHODS

This study involved the creation of a setup comprising four cameras for a 360° photoshoot of human subjects to calibrate and test the developed measurement setup for capturing photos of human subjects and compare the results with manual measurements.

RESULTS

Ten different body dimensions were measured using the setup. There was a significant correlation between the manual and photogrammetric measurement methods (0.943 <  r <  0.997). The highest absolute error recorded was 1.87% .

CONCLUSION

The photogrammetric method for collecting anthropometric data is a reliable substitute for manual measurements across diverse populations. The results indicate that this low-cost approach is highly precise and reliable, with strong correlation to manual measurements. Multiview photogrammetry proves effective for individuals of various body shapes, making it a versatile option for data collection.

Use of a Novel Deep Learning Open-Source Model for Quantification of Ki-67 in Breast Cancer Patients in Pakistan: A Comparative Study between the Manual and Automated Methods

Introduction: Breast cancer is the most common cancer in women; its early detection plays a crucial role in improving patient outcomes. Ki-67 is a biomarker commonly used for evaluating the proliferation of cancer cells in breast cancer patients. The quantification of Ki-67 has traditionally been performed by pathologists through a manual examination of tissue samples, which can be time-consuming and subject to inter- and intra-observer variability. In this study, we used a novel deep learning model to quantify Ki-67 in breast cancer in digital images prepared by a microscope-attached camera. Objective: To compare the automated detection of Ki-67 with the manual eyeball/hotspot method. Place and duration of study: This descriptive, cross-sectional study was conducted at the Jinnah Sindh Medical University. Glass slides of diagnosed cases of breast cancer were obtained from the Aga Khan University Hospital after receiving ethical approval. The duration of the study was one month. Methodology: We prepared 140 digital images stained with the Ki-67 antibody using a microscope-attached camera at 10×. An expert pathologist (P1) evaluated the Ki-67 index of the hotspot fields using the eyeball method. The images were uploaded to the DeepLiif software to detect the exact percentage of Ki-67 positive cells. SPSS version 24 was used for data analysis. Diagnostic accuracy was also calculated by other pathologists (P2, P3) and by AI using a Ki-67 cut-off score of 20 and taking P1 as the gold standard. Results: The manual and automated scoring methods showed a strong positive correlation as the kappa coefficient was significant. The p value was <0.001. The highest diagnostic accuracy, i.e., 95%, taking P1 as gold standard, was found for AI, compared to pathologists P2 and P3. Conclusions: Use of quantification-based deep learning models can make the work of pathologists easier and more reproducible. Our study is one of the earliest studies in this field. More studies with larger sample sizes are needed in future to develop a cohort.

Computerized Tomography Texture Analysis in the Differential Diagnosis of Intracranial Epidermoid and Arachnoid Cysts

PURPOSE

This study evaluated the differences between arachnoid and epidermoid cysts in computerized tomography (CT) texture analysis (TA).

MATERIAL AND METHODS

The study included 12 patients with intracranial epidermoid cysts and 26 patients with intracranial arachnoid cysts who were diagnosed with diffusion-weighted magnetic resonance imaging (DW-MRI) and who had undergone an unenhanced CT examination before treatment. The LIFEx application software was used to obtain texture features. Eighty-two texture features from 38 lesions were automatically calculated for each lesion. The Shapiro-Wilk test was used to test the normality of the scores, and the Mann-Whitney U Test was used to test the difference between the groups. Receiver operating characteristic (ROC) curves and multivariate logistic regression modeling examined the parameters' diagnostic performances.

RESULTS

The median age of the patients was 53 years (range: 19-88 years). Eighty-two texture parameters were evaluated in the first order: gray-level co-occurrence matrix (GLCM), gray-level run length matrix (GLRLM), neighbor gray-tone difference matrix (NGTDM), and gray-level size zone matrix (GLSZM) groups. There was a statistically significant difference between the arachnoid cyst and the epidermoid cyst in the variables of compacity, compactness 1, compactness 2, sphericity, asphericity, sum average, coarseness, and low gray-level zone (p<0.05). According to the multiple logistic regression model, it was determined that the sum average in the GLCM group (B=-0.11; p=0.015), coarseness (B= 869.5; p=0.044) in the NGTDM group, and morphological sphericity (B=24.18; p=0.047) were the radiomics variables that increased the probability of epidermoid diagnosis. According to the classification table of the model, the sensitivity rate was found to be 83%, and the specificity rate was found to be 96%. Therefore, the probability of accurate model classification was 92%.

CONCLUSION

CT TA is a method that can be applied with high diagnostic accuracy in the differential diagnosis of intracranial epidermoid and arachnoid cysts, especially in patients who cannot undergo an MRI examination.

Comparing Magnetic Resonance Fingerprinting (MRF) and the MAGiC Sequence for Simultaneous T1 and T2 Quantitative Measurements in the Female Pelvis: A Prospective Study

The aim of this study was to explore the potential of magnetic resonance fingerprinting (MRF), an emerging quantitative MRI technique, in measuring relaxation values of female pelvic tissues compared to the conventional magnetic resonance image compilation (MAGiC) sequence. The study included 32 female patients who underwent routine pelvic MRI exams using anterior and posterior array coils on a 3T clinical scanner. Our findings demonstrated significant correlations between MRF and MAGiC measured T1 and T2 values (p < 0.0001) for various pelvic tissues, including ilium, femoral head, gluteus, obturator, iliopsoas, erector spinae, uterus, cervix, and cutaneous fat. The tissue contrasts generated from conventional MRI and synthetic MRF also showed agreement in bone, muscle, and uterus for both T1-weighted and T2-weighted images. This study highlights the strengths of MRF in providing simultaneous T1 and T2 mapping. MRF offers distinct tissue contrast and has the potential for accurate diagnosis of female pelvic diseases, including tumors, fibroids, endometriosis, and pelvic inflammatory disease. Additionally, MRF shows promise in monitoring disease progression or treatment response. Overall, the study demonstrates the potential of MRF in the field of female pelvic organ imaging and suggests that it could be a valuable addition to the clinical practice of pelvic MRI exams. Further research is needed to establish the clinical utility of MRF and to develop standardized protocols for its implementation in clinical practice.

Suboptimal Chest Radiography and Artificial Intelligence: The Problem and the Solution

Chest radiographs (CXR) are the most performed imaging tests and rank high among the radiographic exams with suboptimal quality and high rejection rates. Suboptimal CXRs can cause delays in patient care and pitfalls in radiographic interpretation, given their ubiquitous use in the diagnosis and management of acute and chronic ailments. Suboptimal CXRs can also compound and lead to high inter-radiologist variations in CXR interpretation. While advances in radiography with transitions to computerized and digital radiography have reduced the prevalence of suboptimal exams, the problem persists. Advances in machine learning and artificial intelligence (AI), particularly in the radiographic acquisition, triage, and interpretation of CXRs, could offer a plausible solution for suboptimal CXRs. We review the literature on suboptimal CXRs and the potential use of AI to help reduce the prevalence of suboptimal CXRs.

Assessment of first-trimester utero-placental vascular morphology by 3D power Doppler ultrasound image analysis using a skeletonization algorithm: the Rotterdam Periconception Cohort

STUDY QUESTION

Can three-dimensional (3D) Power Doppler (PD) ultrasound and a skeletonization algorithm be used to assess first-trimester development of the utero-placental vascular morphology?

SUMMARY ANSWER

The application of 3D PD ultrasonography and a skeletonization algorithm facilitates morphologic assessment of utero-placental vascular development in the first trimester and reveals less advanced vascular morphologic development in pregnancies with placenta-related complications than in pregnancies without placenta-related complications.

WHAT IS KNOWN ALREADY

Suboptimal development of the utero-placental vasculature is one of the main contributors to the periconceptional origin of placenta-related complications. The nature and attribution of aberrant vascular structure and branching patterns remain unclear, as validated markers monitoring first-trimester utero-placental vascular morphologic development are lacking.

STUDY DESIGN, SIZE, DURATION

In this prospective observational cohort, 214 ongoing pregnancies were included before 10 weeks gestational age (GA) at a tertiary hospital between January 2017 and July 2018, as a subcohort of the ongoing Rotterdam Periconception Cohort study.

PARTICIPANTS/MATERIALS, SETTING, METHODS

By combining 3D PD ultrasonography and virtual reality, utero-placental vascular volume (uPVV) measurements were obtained at 7, 9 and 11 weeks GA. A skeletonization algorithm was applied to the uPVV measurements to generate the utero-placental vascular skeleton (uPVS), a network-like structure containing morphologic characteristics of the vasculature. Quantification of vascular morphology was performed by assigning a morphologic characteristic to each voxel in the uPVS (end-, vessel-, bifurcation- or crossing-point) and calculating total vascular network length. A Mann–Whitney U test was performed to investigate differences in morphologic development of the first-trimester utero-placental vasculature between pregnancies with and without placenta-related complications. Linear mixed models were used to estimate trajectories of the morphologic characteristics in the first trimester.

MAIN RESULTS AND THE ROLE OF CHANCE

All morphologic characteristics of the utero-placental vasculature increased significantly in the first trimester (P < 0.005). In pregnancies with placenta-related complications (n = 54), utero-placental vascular branching was significantly less advanced at 9 weeks GA (vessel points P = 0.040, bifurcation points P = 0.050, crossing points P = 0.020, total network length P = 0.023). Morphologic growth trajectories remained similar after adjustment for parity, conception mode, foetal sex and occurrence of placenta-related complications.

LIMITATIONS, REASONS FOR CAUTION

The tertiary setting of this prospective observational study provides high internal, but possibly limited external, validity. Extrapolation of the study’s findings should therefore be addressed with caution.

WIDER IMPLICATIONS OF THE FINDINGS

The uPVS enables assessment of morphologic development of the first-trimester utero-placental vasculature. Further investigation of this innovative methodology needs to determine its added value for the assessment of (patho-) physiological utero-placental vascular development.

STUDY FUNDING/COMPETING INTEREST(S)

This research was funded by the Department of Obstetrics and Gynecology of the Erasmus MC, University Medical Centre, Rotterdam, The Netherlands. There are no conflicts of interest.

TRIAL REGISTRATION NUMBER

Registered at the Dutch Trial Register (NTR6854).

Towards a Visualizable, De-identified Synthetic Biomarker of Human Movement Disorders

Human motion analysis has been a common thread across modern and early medicine. While medicine evolves, analysis of movement disorders is mostly based on clinical presentation and trained observers making subjective assessments using clinical rating scales. Currently, the field of computer vision has seen exponential growth and successful medical applications. While this has been the case, neurology, for the most part, has not embraced digital movement analysis. There are many reasons for this including: the limited size of labeled datasets, accuracy and nontransparent nature of neural networks, and potential legal and ethical concerns. We hypothesize that a number of opportunities are made available by advancements in computer vision that will enable digitization of human form, movements, and will represent them synthetically in 3D. Representing human movements within synthetic body models will potentially pave the way towards objective standardized digital movement disorder diagnosis and building sharable open-source datasets from such processed videos. We provide a perspective of this emerging field and describe how clinicians and computer scientists can navigate this new space. Such digital movement capturing methods will be important for both machine learning-based diagnosis and computer vision-aided clinical assessment. It would also supplement face-to-face clinical visits and be used for longitudinal monitoring and remote diagnosis.

Glioblastoma Surgery Imaging-Reporting and Data System: Validation and Performance of the Automated Segmentation Task

For patients with presumed glioblastoma, essential tumor characteristics are determined from preoperative MR images to optimize the treatment strategy. This procedure is time-consuming and subjective, if performed by crude eyeballing or manually. The standardized GSI-RADS aims to provide neurosurgeons with automatic tumor segmentations to extract tumor features rapidly and objectively. In this study, we improved automatic tumor segmentation and compared the agreement with manual raters, describe the technical details of the different components of GSI-RADS, and determined their speed. Two recent neural network architectures were considered for the segmentation task: nnU-Net and AGU-Net. Two preprocessing schemes were introduced to investigate the tradeoff between performance and processing speed. A summarized description of the tumor feature extraction and standardized reporting process is included. The trained architectures for automatic segmentation and the code for computing the standardized report are distributed as open-source and as open-access software. Validation studies were performed on a dataset of 1594 gadolinium-enhanced T1-weighted MRI volumes from 13 hospitals and 293 T1-weighted MRI volumes from the BraTS challenge. The glioblastoma tumor core segmentation reached a Dice score slightly below 90%, a patientwise F1-score close to 99%, and a 95th percentile Hausdorff distance slightly below 4.0 mm on average with either architecture and the heavy preprocessing scheme. A patient MRI volume can be segmented in less than one minute, and a standardized report can be generated in up to five minutes. The proposed GSI-RADS software showed robust performance on a large collection of MRI volumes from various hospitals and generated results within a reasonable runtime.

Determination of the Gender-Related Differences on Native Femoral Anatomy Using Three-Dimensional Computerized Tomography Models in Caucasian Population

Introduction Three-dimensional (3D) modelling, which has an increasing interest in the literature, could help surgeons to better understand the lesions by visualizing the real anatomical features compared to plain radiographs and two-dimensional (2D) tomography images. We aimed to evaluate the native femoral anatomical features of Turkish females and males using 3D computed tomography models. Methods We evaluated the right femoral anatomical features of 60 females and 60 males between 31 and 65 years of age creating 3D computerized tomography models. The gender-specific differences of femoral neck inclination and anteversion, femoral mechanical-anatomical axis, anatomical and mechanical lateral distal femoral, medial and lateral proximal femoral angles were measured on 3D femoral anatomical models. Results The mean age of our study groups was 50.6 ± 8.5. We determined a statistically significant difference between gender groups in terms of mean femoral neck anteversion angles (p = 0.009). We observed the retroversion of the femoral neck in 12 adults (10%). The mean values of femoral neck inclination, femoral mechanical - anatomical angle, anatomical and mechanical lateral distal femoral angles, medial and lateral proximal femoral angles did not differ any statistical significance between gender groups. Conclusion Although the anatomical angle measurements except femoral neck anteversion, did not differ significantly between gender groups of our study, there were differences between mean anatomical angles when compared to other studies in the literature, which investigate the different races or Caucasian population. Through 3D anatomical data, more compatible implants, prosthesis or biomaterials can be produced by determining gender and race-specific anatomical differences.

Novel Three-Dimensional Bladder Reconstruction Model from B-Mode Ultrasound Image to Improve the Accuracy of Bladder Volume Measurement

Traditional bladder volume measurement from B-mode (two-dimensional) ultrasound has been found to produce inaccurate results, and thus in this work we aim to improve the accuracy of measurement from B-mode ultrasound. A total of 75 electronic medical records including ultrasonic images were reviewed retrospectively from 64 patients. We put forward a novel bladder volume measurement method, in which a three-dimensional (3D) reconstruction model was established from conventional two-dimensional (2D) ultrasonic images to estimate the bladder volume. The differences and relationships were analyzed among the actual volume, the traditional estimated volume, and the new reconstruction model estimated volume. We also compared the data in different volume groups from small volume to high volume. The mean actual volume is 531.8 mL and the standard deviation is 268.7 mL; the mean percentage error of traditional estimation is −28%. In our new bladder measurement method, the mean percentage error is −10.18% (N = 2), −4.72% (N = 3), −0.33% (N = 4), and 2.58% (N = 5). There is no significant difference between the actual volume and our new bladder measurement method (N = 4) in all data or the divided four groups. The estimated volumes from the traditional method or our new method are highly correlated with the actual volume. Our data show that the three-dimensional bladder reconstruction model provides an accurate measurement from conventional B-mode ultrasonic images compared with the traditional method. The accuracy is seen across different groups of volume, and thus we can conclude that this is a reliable and economical volume measurement model that can be applied in general software or in apps on mobile devices.

Glioblastoma Surgery Imaging-Reporting and Data System: Standardized Reporting of Tumor Volume, Location, and Resectability Based on Automated Segmentations

Treatment decisions for patients with presumed glioblastoma are based on tumor characteristics available from a preoperative MR scan. Tumor characteristics, including volume, location, and resectability, are often estimated or manually delineated. This process is time consuming and subjective. Hence, comparison across cohorts, trials, or registries are subject to assessment bias. In this study, we propose a standardized Glioblastoma Surgery Imaging Reporting and Data System (GSI-RADS) based on an automated method of tumor segmentation that provides standard reports on tumor features that are potentially relevant for glioblastoma surgery. As clinical validation, we determine the agreement in extracted tumor features between the automated method and the current standard of manual segmentations from routine clinical MR scans before treatment. In an observational consecutive cohort of 1596 adult patients with a first time surgery of a glioblastoma from 13 institutions, we segmented gadolinium-enhanced tumor parts both by a human rater and by an automated algorithm. Tumor features were extracted from segmentations of both methods and compared to assess differences, concordance, and equivalence. The laterality, contralateral infiltration, and the laterality indices were in excellent agreement. The native and normalized tumor volumes had excellent agreement, consistency, and equivalence. Multifocality, but not the number of foci, had good agreement and equivalence. The location profiles of cortical and subcortical structures were in excellent agreement. The expected residual tumor volumes and resectability indices had excellent agreement, consistency, and equivalence. Tumor probability maps were in good agreement. In conclusion, automated segmentations are in excellent agreement with manual segmentations and practically equivalent regarding tumor features that are potentially relevant for neurosurgical purposes. Standard GSI-RADS reports can be generated by open access software.

Soft tissue-based registration of intraoral scan with cone beam computed tomography scan

The aim of this study was to evaluate a novel soft tissue-based method to register an intraoral scan (IOS) with a cone beam computed tomography (CBCT) scan. IOS and CBCT data were obtained from eight dentate patients (mean age 21±2 years; three male, five female) and 14 fully edentulous patients (mean age 56±9 years; eight male, six female). An algorithm was developed to create a soft tissue model of the CBCT scan, which allowed a soft tissue-based registration to be performed with the IOS. First, validation was performed on dentate jaws with registration of the palatal mucosal surface and accuracy evaluation at the level of the teeth. Second, fully edentulous jaws were registered using both the palatal and alveolar crest mucosal surfaces. Distance maps were created to measure the method accuracy. The mean registration error was 0.49±0.26mm for the dentate jaws. Registration of the fully edentulous jaws had a mean error of 0.16±0.08mm at the palate and 0.16±0.05mm at the alveolar crest. In conclusion, the high accuracy of this registration method may allow the digital workflow to be optimized when no teeth are available to perform a regular registration procedure.

Comparison of Ki-67 Labeling Index Patterns of Diffuse Large B-Cell Lymphomas and Burkitt Lymphomas Using Image Analysis: A Multicenter Study

Diffuse large B-cell lymphoma (DLBCL) is the most common high-grade B-cell lymphoma found in Korea; it manifests with a variety of cellular morphologies and a high proliferation index. It is difficult to differentiate between DLBCL and Burkitt lymphoma (BL) based on immunohistochemistry, histology, and Epstein-Barr virus infection status owing to the overlap in findings. In this study, we performed comparative morphometric analysis to understand the proportional difference in Ki-67 staining between DLBCL and BL. We analyzed Ki-67-stained slides of 103 DLBCLs and 29 BLs that were pathologically confirmed using a three-tier classification system (negative, 1+, 2+, and 3+) to compare Ki-67 expression between BL and activated B-cell and germinal center B-cell subtypes of DLBCL and DLBCL with high proliferation indices (>90% of 2+ and 3+ cells). Patients with DLBCL were older than those with BL (62.1 versus 51.0 years). The number and proportion of negative cells (passenger and true negative cells) were significantly lower in BLs than those in DLBCLs (337.4, 5.9% versus 690.3, 12.4%). The number and proportion of 3+ cells were significantly higher in BLs than those in DLBCLs (5213.6, 96.3% versus 3132.4, 62.0%). BLs and DLBCLs with a high proliferation index showed similar results as those between BLs and overall DLBCLs. We were able to differentiate BLs and DLBCLs with 98.1% sensitivity and 100.0% specificity using an optimal cut-off of 97.9% of 2+/3+ Ki-67-positive cells. Thus, the Ki-67 labeling index may be a good differential biomarker for DLBCLs and BLs.

Automatic Evaluation of Crown Preparation Using Image Processing Techniques: A Substitute to Faculty Scoring in Dental Education

Background:

This study presents a new and innovative experimental method, including software and its prerequisite instruments, to use image processing techniques for crown preparation analysis.

Method:

A platform was designed and constructed to take images from artificial teeth in different angles and directions and to process and analyze them by the proposed method to evaluate the quality and quantity of crown preparation. For each tooth, two series of images were taken from the artificial teeth before and after preparation, and image series were registered by two semi-automated and automated methods to transform them into one coordinate system. Region of interest was segmented by user interaction, and tooth region was segmented by substeps such as transformation to hue, saturation, and value color space, edge detection, morphology operations, and contour extraction. Finally, the amount and angle of crown preparation were computed and compared with standard measures to evaluate the quality of crown preparation. The proposed method was applied to a local dataset collected from Isfahan University of Medical Sciences.

Results:

Difference between the angle of crown preparation computed by the proposed method and that of the experts showed a mean absolute error of 7.17°. The correlation between the segmented regions by the proposed method and those of the experts was also evaluated by the Intersection over Union (IOU) criterion. The best and worst performances achieved in cases by IOU were 0.94 and 0.76, respectively. Finally, the segmentation results of the proposed method indicated an average IOU of 0.89 in all images.

Conclusion:

Students can use this method as an assessment tool in preclinical tooth preparation to compare their crown work with standard parameters.

Combination of Deep Learning-Based Denoising and Iterative Reconstruction for Ultra-Low-Dose CT of the Chest: Image Quality and Lung-RADS Evaluation

OBJECTIVE. The objective of our study was to assess the effect of the combination of deep learning-based denoising (DLD) and iterative reconstruction (IR) on image quality and Lung Imaging Reporting and Data System (Lung-RADS) evaluation on chest ultra-low-dose CT (ULDCT). MATERIALS AND METHODS. Forty-one patients with 252 nodules were evaluated retrospectively. All patients underwent ULDCT (mean ± SD, 0.19 ± 0.01 mSv) and standard-dose CT (SDCT) (6.46 ± 2.28 mSv). ULDCT images were reconstructed using hybrid iterative reconstruction (HIR) and model-based iterative reconstruction (MBIR), and they were postprocessed using DLD (i.e., HIR-DLD and MBIR-DLD). SDCT images were reconstructed using filtered back projection. Three independent radiologists subjectively evaluated HIR, HIR-DLD, MBIR, and MBIR-DLD images on a 5-point scale in terms of noise, streak artifact, nodule edge, clarity of small vessels, homogeneity of the normal lung parenchyma, and overall image quality. Two radiologists independently evaluated the nodules according to Lung-RADS using HIR, MBIR, HIR-DLD, and MBIR-DLD ULDCT images and SDCT images. The median scores for subjective analysis were analyzed using Wilcoxon signed rank test with Bonferroni correction. Intraobserver agreement for Lung-RADS category between ULDCT and SDCT was evaluated using the weighted kappa coefficient. RESULTS. In the subjective analysis, ULDCT with DLD showed significantly better scores than did ULDCT without DLD (p < 0.001), and MBIR-DLD showed the best scores among the ULDCT images (p < 0.001) for all items. In the Lung-RADS evaluation, HIR showed fair or moderate agreement (reader 1 and reader 2: κw = 0.46 and 0.32, respectively); MBIR, moderate or good agreement (κw = 0.68 and 0.57); HIR-DLD, moderate agreement (κw = 0.53 and 0.48); and MBIR-DLD, good agreement (κw = 0.70 and 0.72). CONCLUSION. DLD improved the image quality of both HIR and MBIR on ULDCT. MBIR-DLD was superior to HIR_DLD for image quality and for Lung-RADS evaluation.

Semi-automatic magnetic resonance imaging based orbital fat volumetry: reliability and correlation with computed tomography

Post-processing analysis can provide valuable information for diagnosis and planning of orbital disorders. This cross-sectional study aims to evaluate the reliability of semi-automatic, orbital fat volumetry using magnetic resonance imaging (MRI). Two observers assessed the orbital fat volume using a standard MRI protocol (3T, T1w sequence) in 12 orbits diagnosed with Graves' orbitopathy (GO) and 10 healthy control orbits. MRI and computed tomography (CT) based analysis were compared. Intra-observer variability was good (intraclass correlation coefficient (ICC) 0.88; 95% confidence interval (CI) [0.70, 0.95]) and interobserver agreement was moderate (ICC 0.55; 95% CI [-0.09, 0.81]), which corresponds to a mean percentage difference of 1.3% and 17.9% of the total orbital fat volume. Mean differences between MRI and CT measurements were, respectively, 1.1 cm³ (P= 0.064, 95% CI [-0.20, 2.43]) and 1.4 cm³ (P=0.016, 95% CI [0.21, 2.56]) for the control and the GO group. MRI volumetry was strongly correlated with CT (Pearson's r= 0.7, P<0.001). We conclude that orbital fat volumetry is feasible with a semi-automatic segmentation procedure and standard MRI protocol. Correlation with CT volumetry is good, but considerable bias may derive from observer variability and these errors should be taken into account for the purpose of volumetric analysis. Better definition of error sources may increase measurement accuracy.

A Novel Solution Based on Scale Invariant Feature Transform Descriptors and Deep Learning for the Detection of Suspicious Regions in Mammogram Images

BACKGROUND

Deep learning methods have become popular for their high-performance rate in the classification and detection of events in computer vision tasks. Transfer learning paradigm is widely adopted to apply pretrained convolutional neural network (CNN) on medical domains overcoming the problem of the scarcity of public datasets. Some investigations to assess transfer learning knowledge inference abilities in the context of mammogram screening and possible combinations with unsupervised techniques are in progress.

METHODS

We propose a novel technique for the detection of suspicious regions in mammograms that consist of the combination of two approaches based on scale invariant feature transform (SIFT) keypoints and transfer learning with pretrained CNNs such as PyramidNet and AlexNet fine-tuned on digital mammograms generated by different mammography devices. Preprocessing, feature extraction, and selection steps characterize the SIFT-based method, while the deep learning network validates the candidate suspicious regions detected by the SIFT method.

RESULTS

The experiments conducted on both mini-MIAS dataset and our new public dataset Suspicious Region Detection on Mammogram from PP (SuReMaPP) of 384 digital mammograms exhibit high performances compared to several state-of-the-art methods. Our solution reaches 98% of sensitivity and 90% of specificity on SuReMaPP and 94% of sensitivity and 91% of specificity on mini-MIAS.

CONCLUSIONS

The experimental sessions conducted so far prompt us to further investigate the powerfulness of transfer learning over different CNNs and possible combinations with unsupervised techniques. Transfer learning performances' accuracy may decrease when the training and testing images come out from mammography devices with different properties.

Relationship between CNR and visibility of anatomical structures of cone-beam computed tomography images under different exposure parameters

OBJECTIVES

The purpose of this study was to investigate the image quality of cone beam CT (CBCT) under different exposure parameters and the relationship between contrast-to-noise and visibility of eight anatomical structures.

METHODS

CBCT images for the evaluation of subjective image quality were acquired on an anthropopathic phantom containing a human skeleton embedded in soft tissue equivalent materials using 25 exposure protocols. Visibility of eight anatomical structures was evaluated by five independent observers. Using the SEDENTEXCT IQ Image Quality phantom, the contrast-to-noise ratio (CNR) was calculated by ImageJ software.

RESULTS

A reduction on the visibility of anatomical structures was seen under lower exposure parameters. However, for 84% of the protocols, visibility of anatomical structures remained acceptable even under some lower parameter settings. As CNR increased, the visibility of anatomical structures also increased correspondingly. A change point could be found in the CNR interval 29.42-36.51 after which the visibility of anatomical structures no longer increases with the increase of CNR.

CONCLUSIONS

Although CNR decrease under a lower exposure parameter, the image quality often remained acceptable at exposure levels below the manufacture's recommended settings. It is possible to standardize subjective image quality by physical factors. Currently, it is not possible to predetermine a change point CNR value due to different CBCT machine and variation of diagnostic tasks.

Increased pulmonary artery diameter is associated with reduced FEV1 in former World Trade Center workers

RATIONALE

Occupational exposures at the WTC site after September 11, 2001 have been associated with several presumably inflammatory lower airway diseases. Pulmonary arterial enlargement, as suggested by an increased ratio of the diameter of the pulmonary artery to the diameter of the aorta (PAAr) has been reported as a computed tomographic (CT) scan marker of adverse respiratory health outcomes, including WTC-related disease. In this study, we sought to utilize a novel quantitative CT (QCT) measurement of PAAr to test the hypothesis that an increased ratio is associated with FEV₁ below each subject's statistically determined lower limit of normal (FEV₁  < LLN).

METHODS

In a group of 1,180 WTC workers and volunteers, we examined whether FEV₁  < LLN was associated with an increased QCT-measured PAAr, adjusting for previously identified important covariates.

RESULTS

Unadjusted analyses showed a statistically significant association of FEV₁  < LLN with PAAr (35.3% vs 24.7%, P = 0.0001), as well as with height, body mass index, early arrival at the WTC disaster site, shorter WTC exposure duration, post-traumatic stress disorder checklist (PCL) score, wall area percent and evidence of bronchodilator response. The multivariate logistic regression model confirmed the association of FEV₁  < LLN with PAAr (OR 1.63, 95% CI 1.21, 2.20, P = 0.0015) and all the unadjusted associations, except for PCL score.

CONCLUSIONS

In WTC workers, FEV₁  < LLN is associated with elevated PAAr which, although likely multifactorial, may be related to distal vasculopathy, as has been hypothesized for chronic obstructive pulmonary disease.

Iterative scatter correction for grid-less skeletal radiography allows improved image quality equal to an antiscatter grid in adjunct with dose reduction: a visual grading study of 20 body donors

BACKGROUND

Iterative scatter correction (ISC) is a new technique applicable to plain radiography; comparable to iterative reconstruction for computed tomography, it promises dose reduction and image quality improvement. ISC for bedside chest X-rays has been applied and evaluated for some time and has recently been commercially offered for plain skeletal radiography.

PURPOSE

To analyze the potential of ISC for plain skeletal radiography with regard to image quality improvement, dose reduction, and replacement for an antiscatter grid.

MATERIAL AND METHODS

A total of 385 radiographs with different imaging protocols of the pelvis and cervical spine were acquired from 20 body donors. Radiographs were rated by four radiologists. Ratings were analyzed with visual grading characteristics (VGC) analysis. The area under the VGC curve was used as a measure of difference in image quality.

RESULTS

Without ISC, the grid-less images were rated significantly worse than their grid-based counterparts (0.389, P = 0.005); adding ISC made image quality equal (0.498; P = 0.963). In grid-less imaging, reduction of dose by 50% led to significant image quality impairment (0.415, P = 0.001); this was fully counterbalanced when ISC was added (0.512; P = 0.588).

CONCLUSION

ISC for plain skeletal radiography has the ability to replace the antiscatter grid without image quality impairment, to improve image quality in grid-less imaging, and to reduce patient radiation dose by 50% without substantial loss in image quality.

Diffusion-Weighted MRI of Breast Cancer: Improved Lesion Visibility and Image Quality Using Synthetic b-Values

BACKGROUND

Diffusion-weighted imaging (DWI) is an MRI technique with the potential to serve as an unenhanced breast cancer detection tool. Synthetic b-values produce images with high diffusion weighting to suppress residual background signal, while avoiding additional measurement times and reducing artifacts.

PURPOSE

To compare acquired DWI images (at b = 850 s/mm² ) and different synthetic b-values (at b = 1000-2000 s/mm² ) in terms of lesion visibility, image quality, and tumor-to-tissue contrast in patients with malignant breast tumors.

STUDY TYPE

Retrospective.

POPULATION

Fifty-three females with malignant breast lesions.

FIELD STRENGTH/SEQUENCE

T₂ w, DWI EPI with STIR fat-suppression, and dynamic contrast-enhanced T₁ w at 3T.

ASSESSMENT

From acquired images using b-values of 50 and 850 s/mm² , synthetic images were calculated at b = 1000, 1200, 1400, 1600, 1800, and 2000 s/mm² . Four readers independently rated image quality, lesion visibility, preferred b-value, as well as the lowest and highest b-value, over the range of b-values tested, to provide a diagnostic image.

STATISTICAL TESTS

Medians and mean ranks were calculated and compared using the Friedman test and Wilcoxon signed-rank test. Reproducibility was analyzed by intraclass correlation (ICC), Fleiss, and Cohen's κ.

RESULTS

Relative signal-to-noise and contrast-to-noise ratios decreased with increasing b-values, while the signal-intensity ratio between tumor and tissue increased significantly (P < 0.001). Intermediate b-values (1200-1800 s/mm² ) were rated best concerning image quality and lesion visibility; the preferred b-value mostly lay at 1200-1600 s/mm² . Lowest and highest acceptable b-values were 850 s/mm² and 2000 s/mm² . Interreader agreement was moderate to high concerning image quality (ICC: 0.50-0.67) and lesion visibility (0.70-0.93), but poor concerning preferred and acceptable b-values (κ = 0.032-0.446).

DATA CONCLUSION

Synthetically increased b-values may be a way to improve tumor-to-tissue contrast, lesion visibility, and image quality of breast DWI, while avoiding the disadvantages of performing DWI at very high b-values.

LEVEL OF EVIDENCE

3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1754-1761.

The role of the subchondral layer in osteonecrosis of the femoral head: analysis based on HR-QCT in comparison to MRI findings

BACKGROUND

Non-traumatic avascular osteonecrosis of the femoral head (ONFH) is a severe disease causing destruction of the hip joint, often necessitating total hip arthroplasty (THA) even in young patients. Magnetic resonance imaging (MRI) is commonly used for diagnosis of ONFH, but provides limited insight into the subchondral bone microstructure.

PURPOSE

To analyze routine MRI findings in comparison to high-resolution quantitative computed tomography (HR-QCT) with a special focus on the subchondral layer and to estimate the importance of differences determining the indication for THA.

MATERIAL AND METHODS

Twelve patients with ONFH were included before THA. Preoperative MRI and HR-QCT of the retrieved femoral heads were aligned using a registration algorithm. Pathological findings and trabecular bone parameters in matched areas were analyzed by two readers. McNemar, marginal homogeneity test, and Pearson's correlation coefficient were used for comparison.

RESULTS

Subchondral delamination was found in nine cases on HR-QCT, but missed or underestimated in all but one case on MRI ( P = 0.016). Chondral discontinuity was found in all cases on HR-QCT and in two cases on MRI ( P = 0.016). Areas of complete bone resorption on HR-QCT were linked to high signal intensity on 3D gradient-echo MRI sequences with water-selective excitation, while there was no correlation between trabecular bone parameters and MRI signal intensities in other areas ( P = 0.304).

CONCLUSION

Subchondral delamination, subchondral resorption, and chondral discontinuity are found frequently in advanced stages of ONFH. These lesions tend to be underestimated on conventional MRI. Our results support the importance of CT imaging in the evaluation of ONFH.

Multimodal quantitative magnetic resonance imaging analysis with individualized postprocessing in patients with drug-resistant focal epilepsy and conventional visual inspection negative for epileptogenic lesions

OBJECTIVES

Approximately one-third of candidates for epilepsy surgery have no visible abnormalities on conventional magnetic resonance imaging. This is extremely discouraging, as these patients have a less favorable prognosis. We aimed to evaluate the utility of quantitative magnetic resonance imaging in patients with drug-resistant neocortical focal epilepsy and negative imaging.

METHODS

A prospective study including 46 patients evaluated through individualized postprocessing of five quantitative measures: cortical thickness, white and gray matter junction signal, relaxation rate, magnetization transfer ratio, and mean diffusivity. Scalp video-electroencephalography was used to suggest the epileptogenic zone. A volumetric fluid-attenuated inversion recovery sequence was performed to aid visual inspection. A critical assessment of follow-up was also conducted throughout the study.

RESULTS

In the subgroup classified as having an epileptogenic zone, individualized postprocessing detected abnormalities within the region of electroclinical origin in 9.7% to 31.0% of patients. Abnormalities outside the epileptogenic zone were more frequent, up to 51.7%. In five patients initially included with negative imaging, an epileptogenic structural abnormality was identified when a new visual magnetic resonance imaging inspection was guided by information gleaned from postprocessing. In three patients, epileptogenic lesions were detected after visual evaluation with volumetric fluid-attenuated sequence guided by video electroencephalography.

CONCLUSION

Although quantitative magnetic resonance imaging analyses may suggest hidden structural lesions, caution is warranted because of the apparent low specificity of these findings for the epileptogenic zone. Conversely, these methods can be used to prevent visible lesions from being ignored, even in referral centers. In parallel, we need to highlight the positive contribution of the volumetric fluid-attenuated sequence.

Heterogeneous distribution of doublecortin-expressing cells surrounding the rostral migratory stream in the juvenile mouse

In the postnatal mammalian brain, neural stem cells of the ventricular-subventricular zone continue to generate doublecortin (Dcx)-expressing immature neurons. Throughout life, these immature neurons migrate to the olfactory bulb through the rostral migratory stream (RMS). In this study, we investigated the distribution of these putative immature neurons using enhanced green fluorescent protein (EGFP) expression in the area surrounding the RMS of the juvenile Dcx-EGFP mice. Through the combined use of an optical clearing reagent (a 2,2'-thiodiethanol solution) and two-photon microscopy, we visualized three-dimensionally the EGFP-positive cells in the entire RMS and its surroundings. The resulting wide-field and high-definition images along with computational image processing methods developed in this study were used to comprehensively determine the position of the EGFP-positive cells. Our findings revealed that the EGFP-positive cells were heterogeneously distributed in the area surrounding the RMS. In addition, the orientation patterns of the leading process of these cells, which displayed the morphology of migrating immature neurons, differed depending on their location. These novel results provide highly precise morphological information for immature neurons and suggest that a portion of immature neurons may be detached from the RMS and migrate in various directions.

Machine Learning Algorithms Utilizing Quantitative CT Features May Predict Eventual Onset of Bronchiolitis Obliterans Syndrome After Lung Transplantation

RATIONALE AND OBJECTIVES

Long-term survival after lung transplantation (LTx) is limited by bronchiolitis obliterans syndrome (BOS), defined as a sustained decline in forced expiratory volume in the first second (FEV₁) not explained by other causes. We assessed whether machine learning (ML) utilizing quantitative computed tomography (qCT) metrics can predict eventual development of BOS.

MATERIALS AND METHODS

Paired inspiratory-expiratory CT scans of 71 patients who underwent LTx were analyzed retrospectively (BOS [n = 41] versus non-BOS [n = 30]), using at least two different time points. The BOS cohort experienced a reduction in FEV₁ of >10% compared to baseline FEV₁ post LTx. Multifactor analysis correlated declining FEV₁ with qCT features linked to acute inflammation or BOS onset. Student t test and ML were applied on baseline qCT features to identify lung transplant patients at baseline that eventually developed BOS.

RESULTS

The FEV₁ decline in the BOS cohort correlated with an increase in the lung volume (P = .027) and in the central airway volume at functional residual capacity (P = .018), not observed in non-BOS patients, whereas the non-BOS cohort experienced a decrease in the central airway volume at total lung capacity with declining FEV₁ (P = .039). Twenty-three baseline qCT parameters could significantly distinguish between non-BOS patients and eventual BOS developers (P < .05), whereas no pulmonary function testing parameters could. Using ML methods (support vector machine), we could identify BOS developers at baseline with an accuracy of 85%, using only three qCT parameters.

CONCLUSIONS

ML utilizing qCT could discern distinct mechanisms driving FEV₁ decline in BOS and non-BOS LTx patients and predict eventual onset of BOS. This approach may become useful to optimize management of LTx patients.

Radiomic features from pretreatment biparametric MRI predict prostate cancer biochemical recurrence: Preliminary findings

BACKGROUND

Radiomics or computer-extracted texture features derived from MRI have been shown to help quantitatively characterize prostate cancer (PCa). Radiomics have not been explored depth in the context of predicting biochemical recurrence (BCR) of PCa.

PURPOSE

To identify a set of radiomic features derived from pretreatment biparametric MRI (bpMRI) that may be predictive of PCa BCR.

STUDY TYPE

Retrospective.

SUBJECTS

In all, 120 PCa patients from two institutions, I₁ and I₂ , partitioned into training set D₁ (N = 70) from I₁ and independent validation set D₂ (N = 50) from I₂ . All patients were followed for ≥3 years.

SEQUENCE

3T, T₂ -weighted (T₂ WI) and apparent diffusion coefficient (ADC) maps derived from diffusion-weighted sequences.

ASSESSMENT

PCa regions of interest (ROIs) on T₂ WI were annotated by two experienced radiologists. Radiomic features from bpMRI (T₂ WI and ADC maps) were extracted from the ROIs. A machine-learning classifier (C_BCR ) was trained with the best discriminating set of radiomic features to predict BCR (p_BCR ).

STATISTICAL TESTS

Wilcoxon rank-sum tests with P < 0.05 were considered statistically significant. Differences in BCR-free survival at 3 years using p_BCR was assessed using the Kaplan-Meier method and compared with Gleason Score (GS), PSA, and PIRADS-v2.

RESULTS

Distribution statistics of co-occurrence of local anisotropic gradient orientation (CoLlAGe) and Haralick features from T₂ WI and ADC were associated with BCR (P < 0.05) on D₁ . C_BCR predictions resulted in a mean AUC = 0.84 on D₁ and AUC = 0.73 on D₂ . A significant difference in BCR-free survival between the predicted classes (BCR + and BCR-) was observed (P = 0.02) on D₂ compared to those obtained from GS (P = 0.8), PSA (P = 0.93) and PIRADS-v2 (P = 0.23).

DATA CONCLUSION

Radiomic features from pretreatment bpMRI can be predictive of PCa BCR after therapy and may help identify men who would benefit from adjuvant therapy.

LEVEL OF EVIDENCE

4 Technical Efficacy: Stage 5 J. Magn. Reson. Imaging 2018;48:1626-1636.

A Modified Tri-Exponential Model for Multi-b-value Diffusion-Weighted Imaging: A Method to Detect the Strictly Diffusion-Limited Compartment in Brain

Purpose: To present a new modified tri-exponential model for diffusion-weighted imaging (DWI) to detect the strictly diffusion-limited compartment, and to compare it with the conventional bi- and tri-exponential models.

Methods: Multi-b-value diffusion-weighted imaging (DWI) with 17 b-values up to 8,000 s/mm² were performed on six volunteers. The corrected Akaike information criterions (AICc) and squared predicted errors (SPE) were calculated to compare these three models.

Results: The mean f₀ values were ranging 11.9–18.7% in white matter ROIs and 1.2–2.7% in gray matter ROIs. In all white matter ROIs: the AICcs of the modified tri-exponential model were the lowest (p < 0.05 for five ROIs), indicating the new model has the best fit among these models; the SPEs of the bi-exponential model were the highest (p < 0.05), suggesting the bi-exponential model is unable to predict the signal intensity at ultra-high b-value. The mean ADC_very−slow values were extremely low in white matter (1–7 × 10⁻⁶ mm²/s), but not in gray matter (251–445 × 10⁻⁶ mm²/s), indicating that the conventional tri-exponential model fails to represent a special compartment.

Conclusions: The strictly diffusion-limited compartment may be an important component in white matter. The new model fits better than the other two models, and may provide additional information.

In vivo coronary artery plaque assessment with computed tomography angiography: is there an impact of iterative reconstruction on plaque volume and attenuation metrics?

Background Coronary computed tomography angiography (CTA) allows the evaluation of coronary plaque volume and low attenuation (lipid-rich) component, for plaque vulnerability assessment. Purpose To determine the effect of iterative reconstruction (IR) on coronary plaque volume and composition. Material and Methods Consecutive patients without coronary artery disease were prospectively enrolled for 256-slice CT. Images were reconstructed with both filtered back projection (FBP) and a hybrid IR algorithm (iDose⁴, Philips) levels 1, 3, 5, and 7. Coronary plaques were assessed according to predefined Hounsfield unit (HU) attenuation intervals, for total plaque and HU-interval volumes. Results Fifty-three patients (mean age, 53.6 years) were included. Noise was significantly decreased and signal-to-noise ratio (SNR) / contrast-to-noise (CNR) were both significantly improved at all IR levels in comparison to FBP. Plaque characterization was performed in 41 patients for a total of 125 plaques. Total plaque volume ranged from 104.4 ± 120.7 to 107.4 ± 128.9 mm³ and low attenuation plaque component from 40.5 ± 54.7 to 43.5 ± 58.9 mm³, with no statistically significant differences between all IR levels and FBP ( P = 0.786 and P ≥ 0.078, respectively). Conclusion IR improved image quality. Total and low attenuation plaque volumes were similar using either IR or FBP.

[Application of area percent of positive neuron and relative intensity of staining grey level in the image quantitative analysis of rat brain tissues immunohistochemistry]

OBJECTIVE

The acetylcholine expression in hypothalamus arcuate nucleus is detected and then the images are processed and analyzed. The features of the image quantitative analysis of immunohistochemistry (IHC) with the method combining two parameters of area percent of positive neuron (APPN) and relative intensity of staining grey level (RISGL) were investigated.

METHODS

Samples were the im-munohistochemical slices of acetylcholine(ACh)expression of hypothalamic arcuate nucleus cholinergic neurons in the process of exercise in-duced immunosuppression, which included twelve groups of "0 w, 2 w, 4 w, 6 w" and three groups of "control, immediately after exercise, 3 hours after exercise" in every week. IHC technology was used to detect the ACh expression. The image quantitative analysis of IHC was con-ducted in accordance with the parameters of ACh total area of positive neuron (TAPN), average intensity of staining grey level (AISGL), APPN, RISGL, APPN/RISGL. Then the differences among APPN, RISGL and traditional parameters in the quantitative analysis were com-pared and the advantages were found.

RESULTS

The changes of TAPN and APPN showed almost the same variation. Namely the corresponding significant differences could be found through these two parameters(P < 0.05), but the sensitivity and anti-interference of APPN was higher. The results of AISGL and RISGL were not coincident completely. Furthermore, with the combination of APPN and RISGL, the positive expres-sion could be reflected better than any single parameter.

CONCLUSIONS

The parameters of immunohistochemical image analysis, APPN and RIS-GL, can be reliable and accurate in image quantitative analysis of IHC. The combination of APPN and RISGL can not only reflect the expres-sion of positive neurons, but also help analyze its mechanism, which is better than traditional analysis parameters.

Effects of exposure parameters and slice thickness on detecting clear and unclear mandibular canals using cone beam CT

OBJECTIVES

The purpose of this study was to clarify the effects of exposure parameters and image-processing methods when using CBCT to detect clear and unclear mandibular canals (MCs).

METHODS

24 dry half mandibles were divided into 2 groups with clear and unclear MCs based on a previous CBCT study. Mandibles were scanned using a CBCT system with varying exposure parameters (tube voltages 60 kV, 70 kV and 90 kV; and tube currents 2 mA, 5 mA, 10 mA and 15 mA) to obtain a total of 144 scans. The images were processed with different slice thicknesses using ImageJ software (National Institutes of Health, Bethesda, MD). Five radiologists evaluated the cross-sectional images of the first molar region to detect the MCs. The diagnostic accuracy of varying exposure parameters and image-processing conditions was compared with the area under the curve (Az) in receiver-operating characteristic analysis.

RESULTS

The Az values for clear MCs were higher than those for unclear MCs (p < 0.0001). With increasing exposure voltages and currents, Az values increased, but no significant differences were found with high voltages and currents in clear MCs (p = 1.0000 and p = 0.9340). The Az values of serial images were higher than those of overlaid images (p < 0.0001), and those for thicker slices were higher than those for thinner slices (p < 0.0001).

CONCLUSIONS

Our findings indicate that detection of unclear MCs requires either higher exposure parameters or processing of the images with thicker slices. To detect clear MCs, lower exposure parameters can be used.

Assessment of three methods of geometric image reconstruction for digital subtraction radiography

OBJECTIVES

To evaluate three methods of geometric image reconstruction for digital subtraction radiography (DSR).

METHODS

Digital periapical radiographs were acquired of 24 teeth with the X-ray tube at 6 different geometric configurations of vertical (V) and horizontal (H) angles: V0°H0°, V0°H10°, V10°H0°, V10°H10°, V20°H0° and V20°H10°. All 144 images were registered in pairs (Group V0°H0° + 1 of the 6 groups) 3 times by using the Emago(®) (Oral Diagnostic Systems, Amsterdam, Netherlands) with manual selection and Regeemy with manual and automatic selections. After geometric reconstruction on the two software applications under different modes of selection, all images were subtracted and the standard deviation of grey values was obtained as a measure of image noise. All measurements were repeated after 15 days to evaluate the method error. Values of image noise were statistically analyzed by one-way ANOVA for differences between methods and between projection angles, followed by Tukey's test at a level of significance of 5%.

RESULTS

Significant differences were found between most of the projection angles for the three reconstruction methods. Image subtraction after manual selection-based reconstruction on Regeemy presented the lowest values of image noise, except on group V0°H0°. The groups V10°H0° and V20°H0° were not significantly different between the manual selection-based reconstruction in Regeemy and automatic selection-based reconstruction in Regeemy methods.

CONCLUSIONS

The Regeemy software on manual mode revealed better quality of geometric image reconstruction for DSR than the Regeemy on automatic mode and the Emago on manual mode, when the radiographic images were obtained at V and H angles used in the present investigation.

Automatic detection of osteoporosis based on hybrid genetic swarm fuzzy classifier approaches

OBJECTIVES

This study proposed a new automated screening system based on a hybrid genetic swarm fuzzy (GSF) classifier using digital dental panoramic radiographs to diagnose females with a low bone mineral density (BMD) or osteoporosis.

METHODS

The geometrical attributes of both the mandibular cortical bone and trabecular bone were acquired using previously developed software. Designing an automated system for osteoporosis screening involved partitioning of the input attributes to generate an initial membership function (MF) and a rule set (RS), classification using a fuzzy inference system and optimization of the generated MF and RS using the genetic swarm algorithm. Fivefold cross-validation (5-FCV) was used to estimate the classification accuracy of the hybrid GSF classifier. The performance of the hybrid GSF classifier has been further compared with that of individual genetic algorithm and particle swarm optimization fuzzy classifiers.

RESULTS

Proposed hybrid GSF classifier in identifying low BMD or osteoporosis at the lumbar spine and femoral neck BMD was evaluated. The sensitivity, specificity and accuracy of the hybrid GSF with optimized MF and RS in identifying females with a low BMD were 95.3%, 94.7% and 96.01%, respectively, at the lumbar spine and 99.1%, 98.4% and 98.9%, respectively, at the femoral neck BMD. The diagnostic performance of the proposed system with femoral neck BMD was 0.986 with a confidence interval of 0.942-0.998. The highest mean accuracy using 5-FCV was 97.9% with femoral neck BMD.

CONCLUSIONS

The combination of high accuracy along with its interpretation ability makes this proposed automatic system using hybrid GSF classifier capable of identifying a large proportion of undetected low BMD or osteoporosis at its early stage.

A comparative study of new and current methods for dental micro-CT image denoising

OBJECTIVES

The aim of the current study was to evaluate the application of two advanced noise-reduction algorithms for dental micro-CT images and to implement a comparative analysis of the performance of new and current denoising algorithms.

METHODS

Denoising was performed using gaussian and median filters as the current filtering approaches and the block-matching and three-dimensional (BM3D) method and total variation method as the proposed new filtering techniques. The performance of the denoising methods was evaluated quantitatively using contrast-to-noise ratio (CNR), edge preserving index (EPI) and blurring indexes, as well as qualitatively using the double-stimulus continuous quality scale procedure.

RESULTS

The BM3D method had the best performance with regard to preservation of fine textural features (CNREdge), non-blurring of the whole image (blurring index), the clinical visual score in images with very fine features and the overall visual score for all types of images. On the other hand, the total variation method provided the best results with regard to smoothing of images in texture-free areas (CNRTex-free) and in preserving the edges and borders of image features (EPI).

CONCLUSIONS

The BM3D method is the most reliable technique for denoising dental micro-CT images with very fine textural details, such as shallow enamel lesions, in which the preservation of the texture and fine features is of the greatest importance. On the other hand, the total variation method is the technique of choice for denoising images without very fine textural details in which the clinician or researcher is interested mainly in anatomical features and structural measurements.

Boosting in Nonlinear Regression Models with an Application to DCE-MRI Data

BACKGROUND

For the statistical analysis of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data, compartment models are a commonly used tool. By these models, the observed uptake of contrast agent in some tissue over time is linked to physiologic properties like capillary permeability and blood flow. Up to now, models of different complexity have been used, and it is still unclear which model should be used in which situation. In previous studies, it has been found that for DCE-MRI data, the number of compartments differs for different types of tissue, and that in cancerous tissue, it might actually differ over a region of voxels of one DCE-MR image.

OBJECTIVES

To find the appropriate number of compartments and estimate the parameters of a regression model for each voxel in an DCE-MR image. With that, tumors in an DCE-MR image can be located, and for example therapy success can be assessed.

METHODS

The observed uptake of contrast agent in a voxel of an image of some tissue is described by a concentration time curve. This curve can be modeled using a nonlinear regression model. We present a boosting approach with nonlinear regression as base procedure, which allows us to estimate the number of compartments and the related parameters for each voxel of an DCE-MR image. In addition, a spatially regularized version of this approach is proposed.

RESULTS

With the proposed approach, the number of compartments - and with that the complexity of the model - per voxel is not fixed but data-driven, which allows us to fit models of adequate complexity to the concentration time curves of all voxels. The parameters of the model remain nevertheless interpretable because of the underlying compartment model.

CONCLUSIONS

The proposed boosting approaches outperform all competing methods considered in this paper regarding the correct localization of tumors in DCE-MR images as well as the spatial homogeneity of the estimated number of compartments across the image, and the definition of the tumor edge.

Segmentation and surface reconstruction of a cadaver heart on Mimics software

The Visible Korean research team used Mimics software (Materialise, Leuven, Belgium) for the segmentation and subsequent surface reconstruction of heart structures using information obtained from sectioned images of a cadaver. Twenty-six heart components were outlined in advance on Photoshop (Adobe Systems, San Jose, CA, USA). By use of the Mimics, the outlined images were then browsed along with the vertical planes as well as the 3-dimensional surface models, which were immediately built by piling the images. Erroneous delineation was readily detected and revised until satisfactory heart models were acquired. The surface models and the selected sectioned images in horizontal, coronal, and sagittal planes were inputted into a PDF file, where any combinations of reconstructed constituents could be displayed and rotated by the user. Mimics software accelerated the segmentation and surface reconstruction of heart anatomical structures. Similar benefits hopefully result from various serial images of other organs. The PDF file, and plane and stereoscopic image data are being distributed to others, and should prove valuable for medical students and clinicians.

Optimization of dental CBCT exposures through mAs reduction

OBJECTIVES

To investigate the effect of tube current-exposure time (mAs) reduction on clinical and technical image quality for different CBCT scanners, and to determine preliminary minimally acceptable values for the mAs and contrast-to-noise ratio (CNR) in CBCT.

METHODS

A polymethyl methacrylate (PMMA) phantom and an anthropomorphic skull phantom, containing a human skeleton embedded in polyurethane, were scanned using four CBCT devices, including seven exposure protocols. For all protocols, the mAs was varied within the selectable range. Using the PMMA phantom, the CNRAIR was measured and corrected for voxel size. Eight axial slices and one coronal slice showing various anatomical landmarks were selected for each CBCT scan of the skull phantom. The slices were presented to six dentomaxillofacial radiologists, providing scores for various anatomical and diagnostic parameters.

RESULTS

A hyperbolic relationship was seen between CNRAIR and mAs. Similarly, a gradual reduction in clinical image quality was seen at lower mAs values; however, for several protocols, image quality remained acceptable for a moderate or large mAs reduction compared with the standard exposure setting, depending on the clinical application. The relationship between mAs, CNRAIR and observer scores was different for each CBCT device. Minimally acceptable values for mAs were between 9 and 70, depending on the criterion and clinical application.

CONCLUSIONS

Although noise increased at a lower mAs, clinical image quality often remained acceptable at exposure levels below the manufacturer's recommended setting, for certain patient groups. Currently, it is not possible to determine minimally acceptable values for image quality that are applicable to multiple CBCT models.

Effect of exposure parameters and voxel size on bone structure analysis in CBCT

OBJECTIVE

To evaluate the effect of exposure parameters and voxel size on bone structure analysis in dental CBCT.

METHODS

20 cylindrical bone samples underwent CBCT scanning (3D Accuitomo 170; J. Morita, Kyoto, Japan) using three combinations of tube voltage (kV) and tube current-exposure time product (mAs), corresponding with a CT dose index of 3.4 mGy: 90 kV and 62 mAs, 73 kV and 108.5 mAs, and 64 kV and 155 mAs. Images were reconstructed with a voxel size of 0.080 mm. In addition, the 90 kV scan was reconstructed at voxel sizes of 0.125, 0.160, 0.200, 0.250 and 0.300 mm. The following parameters were measured: bone surface (BS) and bone volume (BV) per total volume (TV), fractal dimension, connectivity density, anisotropy, trabecular thickness (Tb. Th.) and trabecular spacing (Tb. Sp.), structure model index (SMI), plateness, branches, junctions, branch length and triple points.

RESULTS

For most parameters, there was no significant effect of the kV value. For BV/TV, "90 kV" differed significantly from the other kV settings; for SMI, "64 vs 73 kV" was significant. For BS/TV, fractal dimension, connectivity density, branches, junctions and triple points values incrementally decreased at larger voxel sizes, whereas an increase was seen for Tb. Th., Tb. Sp., SMI and branch length. For anisotropy and plateness, no (or little) effect of voxel size was seen; for BV/TV, the effect was inconsistent.

CONCLUSIONS

Most bone structure parameters are not affected by the kV if the radiation dose is constant. Parameters dealing with the trabecular structure are heavily affected by the voxel size.

Comparison of mandibular bone microarchitecture between micro-CT and CBCT images

OBJECTIVES

To compare microarchitecture parameters of bone samples scanned using micro-CT (µCT) to those obtained by using CBCT.

METHODS

A bone biopsy trephine bur (3 × 10 mm) was used to remove 20 cylindrical bone samples from 20 dry hemimandibles. Samples were scanned using µCT (µCT 35; SCANCO Medical, Brüttisellen, Switzerland) with a voxel size of 20 µm and CBCT (3D Accuitomo 170; J. Morita, Kyoto, Japan) with a voxel size of 80 µm. All corresponding sample scans were aligned and cropped. Image analysis was carried out using BoneJ, including the following parameters: skeleton analysis, bone surface per total volume (BS/TV), bone volume per total volume (BV/TV), connectivity density, anisotropy, trabecular thickness and spacing, structure model index, plateness and fractal dimension. Pearson and Spearman correlation coefficients (R) were calculated. CBCT values were then calibrated using the slope of the linear fit with the µCT values. The mean error after calibration was calculated and normalized to the standard deviation of the µCT values.

RESULTS

R-values ranged between 0.05 (plateness) and 0.83 (BS/TV). Correlation was significant for both Spearman and Pearson's R for 8 out of 16 parameters. After calibration, the smallest normalized error was found for BV/TV (0.48). For other parameters, the error range was 0.58-2.10.

CONCLUSIONS

Despite the overall correlation, this study demonstrates the uncertainty associated with using bone microarchitecture parameters on CBCT images. Although clinically relevant parameter ranges are not available, the errors found in this study may be too high for some parameters to be considered for clinical application.

Three-dimensional reconstruction of teeth and jaws based on segmentation of CT images using watershed transformation

The objective of the present study was the development and clinical testing of a three-dimensional (3D) reconstruction method of teeth and a bone tissue of the jaw on the basis of CT images of the maxillofacial region. 3D reconstruction was performed using the specially designed original software based on watershed transformation. Computed tomograms in digital imaging and communications in medicine format obtained on multispiral CT and CBCT scanners were used for creation of 3D models of teeth and the jaws. The processing algorithm is realized in the stepwise threshold image segmentation with the placement of markers in the mode of a multiplanar projection in areas relating to the teeth and a bone tissue. The developed software initially creates coarse 3D models of the entire dentition and the jaw. Then, certain procedures specify the model of the jaw and cut the dentition into separate teeth. The proper selection of the segmentation threshold is very important for CBCT images having a low contrast and high noise level. The developed semi-automatic algorithm of multispiral and cone beam computed tomogram processing allows 3D models of teeth to be created separating them from a bone tissue of the jaws. The software is easy to install in a dentist's workplace, has an intuitive interface and takes little time in processing. The obtained 3D models can be used for solving a wide range of scientific and clinical tasks.

ANTONIA perfusion and stroke. A software tool for the multi-purpose analysis of MR perfusion-weighted datasets and quantitative ischemic stroke assessment

OBJECTIVES

The objective of this work is to present the software tool ANTONIA, which has been developed to facilitate a quantitative analysis of perfusion-weighted MRI (PWI) datasets in general as well as the subsequent multi-parametric analysis of additional datasets for the specific purpose of acute ischemic stroke patient dataset evaluation.

METHODS

Three different methods for the analysis of DSC or DCE PWI datasets are currently implemented in ANTONIA, which can be case-specifically selected based on the study protocol. These methods comprise a curve fitting method as well as a deconvolution-based and deconvolution-free method integrating a previously defined arterial input function. The perfusion analysis is extended for the purpose of acute ischemic stroke analysis by additional methods that enable an automatic atlas-based selection of the arterial input function, an analysis of the perfusion-diffusion and DWI-FLAIR mismatch as well as segmentation-based volumetric analyses.

RESULTS

For reliability evaluation, the described software tool was used by two observers for quantitative analysis of 15 datasets from acute ischemic stroke patients to extract the acute lesion core volume, FLAIR ratio, perfusion-diffusion mismatch volume with manually as well as automatically selected arterial input functions, and follow-up lesion volume. The results of this evaluation revealed that the described software tool leads to highly reproducible results for all parameters if the automatic arterial input function selection method is used.

CONCLUSION

Due to the broad selection of processing methods that are available in the software tool, ANTONIA is especially helpful to support image-based perfusion and acute ischemic stroke research projects.

Lung registration using automatically detected landmarks

OBJECTIVES

Accurate registration of lung CT images is inevitable for numerous clinical applications. Usually, nonlinear intensity-based methods are used. Their accuracy is typically evaluated using corresponding anatomical points (landmarks; e.g. bifurcations of bronchial and vessel trees) annotated by medical experts in the images to register. As image registration can be interpreted as correspondence finding problem, these corresponding landmarks can also be used in feature-based registration techniques. Recently, approaches for automated identification of such landmark correspondences in lung CT images have been presented. In this work, a novel combination of variational nonlinear intensity-based registration with an approach for automated landmark correspondence detection in lung CT pairs is presented and evaluated.

METHODS

The main blocks of the proposed hybrid intensity- and feature-based registration scheme are a two-step landmark correspondence detection and the so-called CoLD (Combining Landmarks and Distance Measures) framework. The landmark correspondence identification starts with feature detection in one image followed by a blockmatching-based transfer of the features to the other image. The established correspondences are used to compute a thin-plate spline (TPS) transformation. Within CoLD, the TPS transformation is improved by minimization of an objective function consisting of a Normalized Gradient Field distance measure and a curvature regularizer; the landmark correspondences are guaranteed to be preserved by optimization on the kernel of the discretized landmark constraints.

RESULTS

Based on ten publicly available end-inspiration/expiration CT scan pairs with anatomical landmark sets annotated by medical experts from the DIR-Lab database, it is shown that the hybrid registration approach is superior in terms of accuracy: The mean distance of expert landmarks is decreased from 8.46 mm before to 1.15 mm after registration, outperforming both the TPS transformation (1.68 mm) and a nonlinear registration without usage of automatically detected landmarks (2.44 mm). The improvement is statistically significant in eight of ten datasets in comparison to TPS and in nine of ten datasets in comparison to the intensity-based registration. Furthermore, CoLD globally estimates the breathing-induced lung volume change well and results in smooth and physiologically plausible motion fields of the lungs.

CONCLUSIONS

We demonstrated that our novel landmark-based registration pipeline outperforms both TPS and the underlying nonlinear intensity-based registration without landmark usage. This highlights the potential of automatic landmark correspondence detection for improvement of lung CT registration accuracy.

Reliability of neuroanatomical measurements in a multisite longitudinal study of youth at risk for psychosis

Multisite longitudinal neuroimaging designs are used to identify differential brain structural change associated with onset or progression of disease. The reliability of neuroanatomical measurements over time and across sites is a crucial aspect of power in such studies. Prior work has found that while within-site reliabilities of neuroanatomical measurements are excellent, between-site reliability is generally more modest. Factors that may increase between-site reliability include standardization of scanner platform and sequence parameters and correction for between-scanner variations in gradient nonlinearities. Factors that may improve both between- and within-site reliability include use of registration algorithms that account for individual differences in cortical patterning and shape. In this study 8 healthy volunteers were scanned twice on successive days at 8 sites participating in the North American Prodrome Longitudinal Study (NAPLS). All sites employed 3 Tesla scanners and standardized acquisition parameters. Site accounted for 2 to 30% of the total variance in neuroanatomical measurements. However, site-related variations were trivial (<1%) among sites using the same scanner model and 12-channel coil or when correcting for between-scanner differences in gradient nonlinearity and scaling. Adjusting for individual differences in sulcal-gyral geometries yielded measurements with greater reliabilities than those obtained using an automated approach. Neuroimaging can be performed across multiple sites at the same level of reliability as at a single site, achieving within- and between-site reliabilities of 0.95 or greater for gray matter density in the majority of voxels in the prefrontal and temporal cortical surfaces as well as for the volumes of most subcortical structures.

A statistical cerebroarterial atlas derived from 700 MRA datasets

OBJECTIVES

The cerebroarterial system is a complex network of arteries that supply the brain cells with vitally important nutrients and oxygen. The inter-individual differences of the cerebral arteries, especially at a finer level, are still not understood sufficiently. The aim of this work is to present a statistical cerebroarterial atlas that can be used to overcome this problem.

METHODS

Overall, 700 Time-of-Flight (TOF) magnetic resonance angiography (MRA) datasets of healthy subjects were used for atlas generation. Therefore, the cerebral arteries were automatically segmented in each dataset and used for a quantification of the vessel diameters. After this, each TOF MRA dataset as well as the corresponding vessel segmentation and vessel diameter dataset were registered to the MNI brain atlas. Finally, the registered datasets were used to calculate a statistical cerebroarterial atlas that incorporates information about the average TOF intensity, probability for a vessel occurrence and mean vessel diameter for each voxel.

RESULTS

Visual analysis revealed that arteries with a diameter as small as 0.5 mm are well represented in the atlas with quantitative values that are within range of anatomical reference values. Moreover, a highly significant strong positive correlation between the vessel diameter and occurrence probability was found. Furthermore, it was shown that an intensity-based automatic segmentation of cerebral vessels can be considerable improved by incorporating the atlas information leading to results within the range of the inter-observer agreement.

CONCLUSION

The presented cerebroarterial atlas seems useful for improving the understanding about normal variations of cerebral arteries, initialization of cerebrovascular segmentation methods and may even lay the foundation for a reliable quantification of subtle morphological vascular changes.

Two-dimensional sectioned images and three-dimensional surface models for learning the anatomy of the female pelvis

In the Visible Korean project, serially sectioned images of the pelvis were made from a female cadaver. Outlines of significant structures in the sectioned images were drawn and stacked to build surface models. To improve the accessibility and informational content of these data, a five-step process was designed and implemented. First, 154 pelvic structures were outlined with additional surface reconstruction to prepare the image data. Second, the sectioned and outlined images (in a browsing software) as well as the surface models (in a PDF file) were placed on the Visible Korean homepage in a readily-accessible format. Third, all image data were visualized with interactive elements to stimulate creative learning. Fourth, two-dimensional (2D) images and three-dimensional (3D) models were superimposed on one another to provide context and spatial information for students viewing these data. Fifth, images were designed such that structure names would be shown when the mouse pointer hovered over the 2D images or the 3D models. The state-of-the-art sectioned images, outlined images, and surface models, arranged and systematized as described in this study, will aid students in understanding the anatomy of female pelvis. The graphic data accompanied by corresponding magnetic resonance images and computed tomographs are expected to promote the production of 3D simulators for clinical practice.

Study of edge detection task in dental panoramic radiographs

The purpose of this study is (1) to introduce a new approach for edge detection in orthopantograms (OPGs) and an improved automatic parameter selector for common edge detectors, (2) to present a comparison between our novel approach with common edge detectors and (3) to provide faster outputs without compromising quality. A new approach for edge detection based on statistical measures was introduced: (1) a set of N edge detection results is calculated from a given input image and a selected type of edge detector, (2) N correspondence maps are constructed from N edge detection results, (3) probabilities and average probabilities are computed, (4) an overall correspondence is evaluated for each correspondence map and (5) the correspondence map providing the best overall correspondence is taken as the result of edge detection procedure. A comparison with common edge detectors (the Roberts, Prewitt, Sobel, Laplacian of the Gaussian and Canny methods) with various parameter settings (304 combinations for each test image) was carried out. The methods were assessed objectively [edge mismatch error (EME), modified Hausdorff distance (MHD) and principal component analysis] and subjectively by experts in dentistry and based on time demands. The suitability of the new approach for edge detection in OPGs was confirmed by experts. The current conventional methods in edge detection in OPGs are inadequate (none of the tested methods reach an EME value or MHD value below 0.1). Our proposed approach for edge detection shows promising potential for its implementation in clinical dentistry. It enhances the accuracy of OPG interpretation and advances diagnosis and treatment planning.

Enabling user-guided segmentation and tracking of surface-labeled cells in time-lapse image sets of living tissues

To study the process of morphogenesis, one often needs to collect and segment time-lapse images of living tissues to accurately track changing cellular morphology. This task typically involves segmenting and tracking tens to hundreds of individual cells over hundreds of image frames, a scale that would certainly benefit from automated routines; however, any automated routine would need to reliably handle a large number of sporadic, and yet typical problems (e.g., illumination inconsistency, photobleaching, rapid cell motions, and drift of focus or of cells moving through the imaging plane). Here, we present a segmentation and cell tracking approach based on the premise that users know their data best-interpreting and using image features that are not accounted for in any a priori algorithm design. We have developed a program, SeedWater Segmenter, that combines a parameter-less and fast automated watershed algorithm with a suite of manual intervention tools that enables users with little to no specialized knowledge of image processing to efficiently segment images with near-perfect accuracy based on simple user interactions.