Development and validation of a new formula for sex estimation based on multislice computed tomographic measurements of maxillary and frontal sinuses among Brazilian adults

Objectives:To evaluate the accuracy of the measurements of the maxillary sinus (MS) and frontal sinus (FS) in sex estimation among Brazilian adults using multislice computed tomography (MCT) and to develop and cross-validate a new formula for sex estimation.Methods:The present cross-sectional research was conducted in two phases: (1) development of a formula on the basis of the measurements of both the sinuses (50 males and 50 females); and (2) validation study (20 males and 20 females). The linear measurements (height, width and diameter) were assessed using the RadiAnt DICOM software. A new formula for sex estimation was developed (multivariate statistical approach) and validated. Receiver operating characteristic curves, area under the curve, sensitivity, specificity, positive and negative predictive values, accuracy and likelihood ratio were estimated.Results:Males displayed higher mean values (width, height and diameter) of the FS and MS (p < 0.05). The MS was a better predictor in sex estimation (males vs females), compared to the FS (accuracy between 61–74% and 58–69%, respectively). The distance between the right and left MS displayed the highest accuracy (74%). The sensitivity, specificity and accuracy of the new formula were 80%, 95.5% and 87.5%, respectively. 63.1% reduction was observed in the number of predictive values for sex estimation (individuals older than 30 years).Conclusions:The present MCT measurements showed a higher accuracy in the estimation of sex in males. The highest accuracy was associated with the distance between the right and left MS. The new formula displayed high precision for sex estimation.     

Sex estimation using foramen magnum measurements,discriminant analyses and artificial neural networks on an eastern Turkish population sample

BackgroundAlthough many studies have been conducted using the foramen magnum for sex estimation, recent findings have indicated that the discriminant and regression models obtained from the foramen magnum may not be reliable. Artificial Neural Networks, was used as a classification technique in sex estimation studies on some other bones, did not used in sex estimation studies on the foramen magnum until now. The aim of this study was sex estimation on an Eastern Turkish population sample using foramen magnum measurements, discriminant analyses and Artificial Neural Networks.MethodologyThe study was performed on the CT images of a total of 720 cases, comprising 360 males and 360 females. For sex estimation, discriminant analysis and Artificial Neural Networks were used.ResultsThe accuracy rate was 86.7% with discriminant analysis and when sex estimation accuracy was determined according to cases with posterior probabilities above 95%, the accuracy ranged from 0% to 33.3%. With the use of the discriminant formulas of 2 other studies, obtained from different Turkish samples, sex could be determined at a rate of 84.6%. Some formulas were found to be unsuccessful in sex estimation. Sex estimation accuracy of 88.2% was achieved with Artificial Neural Networks.ConclusionIn this study, it was found that sex could be determined to some extent with discriminant formulas from other samples from the same population, although some formulas were unsuccessful. With the use of image processing techniques and machine learning algorithms, better results can be obtained in sex estimation.     

Deep-learning for predicting C-shaped canals in mandibular second molars on panoramic radiographs

Objective:The aim of this study was to evaluate the use of a convolutional neural network (CNN) system for predicting C-shaped canals in mandibular second molars on panoramic radiographs.Methods:Panoramic and cone beam CT (CBCT) images obtained from June 2018 to May 2020 were screened and 1020 patients were selected. Our dataset of 2040 sound mandibular second molars comprised 887 C-shaped canals and 1153 non-C-shaped canals. To confirm the presence of a C-shaped canal, CBCT images were analyzed by a radiologist and set as the gold standard. A CNN-based deep-learning model for predicting C-shaped canals was built using Xception. The training and test sets were set to 80 to 20%, respectively. Diagnostic performance was evaluated using accuracy, sensitivity, specificity, and precision. Receiver-operating characteristics (ROC) curves were drawn, and the area under the curve (AUC) values were calculated. Further, gradient-weighted class activation maps (Grad-CAM) were generated to localize the anatomy that contributed to the predictions.Results:The accuracy, sensitivity, specificity, and precision of the CNN model were 95.1, 92.7, 97.0, and 95.9%, respectively. Grad-CAM analysis showed that the CNN model mainly identified root canal shapes converging into the apex to predict the C-shaped canals, while the root furcation was predominantly used for predicting the non-C-shaped canals.Conclusions:The deep-learning system had significant accuracy in predicting C-shaped canals of mandibular second molars on panoramic radiographs.     

The subpubic angle in sex determination: Anthropometric measurements and analyses on Anatolian Caucasians using multidetector computed tomography datasets

Introduction and aimThe investigation of new anatomical criteria and revalidation of existing ones in sex determination for different populations are among main research foci of forensic anthropometry. In that context, the pelvis is the most extensively studied bone. A number of qualitative classifications, dimensional measurements and indices have been proposed for investigative anthropometry and forensic studies. Independent use of these parameters generally provided an accuracy rate of 70–75%. In this study, the accuracy rate of the subpubic angle in sex determination was investigated in living Anatolian Caucasians.Material and methodThe subpubic angle was identified and measured on three-dimensional computed tomographic images of pelves. Data were obtained using 64-detector computed tomography (MDCT) with an isotrophic resolution of 500 μm. The sample included 66 males (41.6 ± 14.9 years of age) and 43 females (41.1 ± 14.2 years of age). Measurements were taken on a dedicated three-dimensional image analysis workstation. The subpubic angle was electronically measured. The technique and methodology was validated on a standard skeletal model. Intraobserver agreement was analyzed with intraclass correlation coefficient, and intraobserver variability was evaluated with technical error of measurement (inter- and intra-observer TEM), relative technical error of measurement (rTEM) and coefficient of reliability (R) measures. The subpubic angle for the study group and for both sexes was reported as minimum–maximum (mean ± SD). Independent-Samples T Test for equality of means was used to determine the difference between the two sexes regarding the subpubic angle. The correlation between the subpubic angle and the age of subjects were using Pearson Correlation Coefficients in males and in females. Logistic regression model was used to classify subjects according to their sex. Receiver operating characteristic curve analysis was performed to determine a cut-off value for further studies and to test the performance of the binary classification test.ResultsIntraclass correlation for the subpubic angle (0.990), TEM (1082), rTEM (1.492), and R (0.990) represented almost complete reliability and accuracy of the measurement method. The subpubic angle was between 48° and 81° (65.9° ± 7.2°) in males and was between 64° and 100° (82.6° ± 7.7°) in females. Statistically significant difference was found between males and females regarding the subpubic angle (p < 0.0001). The subpubic angle was not significantly correlated with age in males (p = 0.953), or in females (r = 0.975). The accuracy of the subpubic angle in sex determination was 90.8%. With a cut-off value of 74°, sensitivity of subpubic angle to detect female phenotype was 88% and its specificity was 95%.ConclusionThe subpubic angle is an accurate parameter in sex determination with high sensitivity and specificity.     

Sex determination from talus in Chinese population by three-dimensional measurement approach

ObjectiveThe determination of sex is an essential part of building the biological profile for unknown human remains. Sex determination from talus in Chinese population has been rarely reported. The aim of this study was to determine sex by discriminant function analysis through talus measurement in Chinese population.Methods48 male and 47 female Chinese northeast subjects were taken in this research. The ankle joints of these subjects were scanned by CT. In total, thirteen indexes were measured through Mimics and Magics software. Length and breadth indexes of total talus, trochlea, talar head, medial and lateral malleolus articular surface were mainly selected. Nine of them were measured through Mimics software. The other four indexes were measured through Magics software. All data were analyzed by independent-samples t-test in SPSS and Stata software. Discriminant function equations were generated for sex determination.ResultsAll the indexes were normally distributed. No significant difference between left and right talus in either males or females was identified (P > 0.05). All results showed significant sexual difference (P < 0.05) except posterior breadth of trochlea. The average accuracy of sex determination ranged from 95.85% to 98.45% in the direct method and 98.95% in the stepwise method.ConclusionsLength indexes showed higher accuracy rate than breadth ones. Length of lateral malleolus articular surface was the best discriminator of sexual dimorphism. Talus was proved effective for sex determination in Chinese population. This study provided a remarkable reference for sex determination in forensic science.     

Forensic bone age estimation of adolescent pelvis X-rays based on two-stage convolutional neural network

In the forensic estimation of bone age, the pelvis is important for identifying the bone age of teenagers. However, studies on this topic remain insufficient as a result of lower accuracy due to the overlapping of pelvic organs in X-ray images. Segmentation networks have been used to automate the location of key pelvic areas and minimize restrictions like doubling images of pelvic organs to increase the accuracy of estimation. This study conducted a retrospective analysis of 2164 pelvis X-ray images of Chinese Han teenagers ranging from 11 to 21 years old. Key areas of the pelvis were detected with a U-Net segmentation network, and the findings were combined with the original X-ray image for regional augmentation. Bone age estimation was conducted with the enhanced and not enhanced pelvis X-ray images by separately using three convolutional neural networks (CNNs). The root mean square errors (RMSE) of the Inception-V3, Inception-ResNet-V2, and VGG19 convolutional neural networks were 0.93 years, 1.12 years, and 1.14 years, respectively, and the mean absolute errors (MAE) of these networks were 0.67 years, 0.77 years, and 0.88 years, respectively. For comparison, a network without segmentation was employed to conduct the estimation, and it was found that the RMSE of the three CNNs above became 1.22 years, 1.25 years, and 1.63 years, respectively, and the MAE became 0.93 years, 0.96 years, and 1.23 years. Bland–Altman plots and attention maps were also generated to provide a visual comparison. The proposed segmentation network can be used to reduce the influence of restrictions like image overlapping of organs and can thus increase the accuracy of pelvic bone age estimation.

     

Deep learning in sex estimation from knee radiographs – A proof-of-concept study utilizing the Terry Anatomical Collection

Although knee measurements yield high classification rates in metric sex estimation, there is a paucity of studies exploring the knee in artificial intelligence-based sexing. This proof-of-concept study aimed to develop deep learning algorithms for sex estimation from radiographs of reconstructed cadaver knee joints belonging to the Terry Anatomical Collection. A total of 199 knee radiographs were obtained from 100 skeletons (46 male and 54 female cadavers; mean age at death 64.2 years, range 50–102 years) whose tibiofemoral joints were reconstructed in standard anatomical position. The AIDeveloper software was used to train, validate, and test neural network architectures in sex estimation based on image classification. Of the explored algorithms, an MhNet-based model reached the highest overall testing accuracy of 90.3%. The model was able to classify all females (100.0%) and most males (78.6%) correctly. These preliminary findings encourage further research on artificial intelligence-based methods in sex estimation from the knee joint. Combining radiographic data with automated and externally validated algorithms may establish valuable tools to be utilized in forensic anthropology.     

The effect of Gaussian noise on pneumonia detection on chest radiographs,using convolutional neural networks

IntroductionChest X-rays (CXR) with under-exposure increase image noise and this may affect convolutional neural network (CNN) performance. This study aimed to train and validate CNNs for classifying pneumonia on CXR as normal or pneumonia acquired at different image noise levels.MethodsThe study used the curated and publicly available “Chest X-Ray Pneumonia” dataset of 5856 AP CXR classified into 1583 normal, 4273 viral and bacterial pneumonia cases. Gaussian noise with zero mean was added to the images, at 5 image noise variance levels, corresponding to decreasing exposure. Each noise-level dataset was split into 80% for training, 10% for validation, and 10% for test data and then classified using custom trained sequential CNN architecture. Six classification tasks were developed for five Gaussian noise levels and the original dataset. Sensitivity, specificity, predictive values and accuracy were used as evaluation performance metrics.ResultsCNN evaluation on the different datasets revealed no performance drop from the original dataset to the five datasets with different noise levels. Sensitivity, specificity and accuracy for the normal datasets were 98.7%, 76.1% and 90.2%. For the five Gaussian noise levels the sensitivity, specificity and accuracy ranged from 96.9% to 98.2%, 94.4%–98.7% and 96.8%–97.6%, respectively. A heat map was used for visual explanation of the CNNs.ConclusionThe CNNs sensitivity maintained, and the specificity increased in distinguishing between normal and pneumonia CXR with the introduction of image noise.Implications for practiceNo performance drops of CNNs in distinguishing cases with and without pneumonia CXR with different Gaussian noise levels was observed. This has potential for decreasing radiation dose to patients or maintaining exposure parameters for patients that require additional radiographs.     

Deep Learning for Automated Classification of Inferior Vena Cava Filter Types on Radiographs

PurposeTo demonstrate the feasibility and evaluate the performance of a deep-learning convolutional neural network (CNN) classification model for automated identification of different types of inferior vena cava (IVC) filters on radiographs.Materials and MethodsIn total, 1,375 cropped radiographic images of 14 types of IVC filters were collected from patients enrolled in a single-center IVC filter registry, with 139 images withheld as a test set and the remainder used to train and validate the classification model. Image brightness, contrast, intensity, and rotation were varied to augment the training set. A 50-layer ResNet architecture with fixed pre-trained weights was trained using a soft margin loss over 50 epochs. The final model was evaluated on the test set.ResultsThe CNN classification model achieved a F1 score of 0.97 (0.92–0.99) for the test set overall and of 1.00 for 10 of 14 individual filter types. Of the 139 test set images, 4 (2.9%) were misidentified, all mistaken for other filter types that appear highly similar. Heat maps elucidated salient features for each filter type that the model used for class prediction.ConclusionsA CNN classification model was successfully developed to identify 14 types of IVC filters on radiographs and demonstrated high performance. Further refinement and testing of the model is necessary before potential real-world application.     

Accurate age estimation from blood samples of Han Chinese individuals using eight high-performance age-related CpG sites

Age-related CpG sites (AR-CpGs) are currently the most promising biomarkers for forensic age estimation. In our previous studies, we first validated the age correlation of seven reported AR-CpGs in blood samples of Chinese Han population. Subsequently, we screened some good age predictors from blood samples of Chinese Han population, and built pyrosequencing-based age prediction models. However, it is still important to select a set of high-performance AR-CpGs in a specific racial group and establish a simple and efficient method for accurate age estimation for forensic purpose. In this study, eight AR-CpGs, namely chr6: 11,044,628 (ELOVL2), cg06639320 (FHL2), chr1: 207,823,723 (C1orf132), cg19283806 (CCDC102B), cg14361627 (KLF14), cg17740900 (SYNE2), cg07553761 (TRIM59), and cg26947034, were selected based on our previous studies, and a multiplex methylation SNaPshot assay was developed to investigate DNA methylation levels at these AR-CpGs in 529 blood samples (aged 2–82 years) from Han Chinese population. All selected CpG sites showed strong age correlation with the correlation coefficient (r) from 0.8363 to 0.9251. Multiple linear regression (MLR) and support vector regression (SVR) age prediction models were simultaneously established to fit change characteristics of DNA methylation levels of eight AR-CpGs with the age in 374 donors’ blood samples. The MLR model enabled age prediction with R² = 0.923, mean absolute error (MAE) = 3.52, while the SVR model enabled age prediction with R² = 0.935, MAE = 2.88. One hundred fifty-five independent samples were used as a validation set to test the two models’ performance, and the prediction MAE for the validation set was 3.71 and 3.34 for the MLR and SVR models, respectively. For the MLR and SVR models, the correct prediction rate at ± 5 years reached a high level of 79.35% and 83.23%, respectively. In general, these statistical parameters indicated that the SVR model outperformed the MLR model in age prediction of the Han Chinese population. In addition, our method provides sufficient sensitivity in forensic applications and allows for 100% efficiency when examining bloodstains kept in room conditions for up to 43 days. These results indicate that our multiplex methylation SNaPshot assay is a reliable, effective, and accurate method for age prediction in blood samples from the Chinese Han population.

     

New models for age estimation and assessment of their accuracy using developing mandibular third molar teeth in a Thai population

Age estimation using developing third molar teeth is considered an important and accurate technique for both clinical and forensic practices. The aims of this study were to establish population-specific reference data, to develop age prediction models using mandibular third molar development, to test the accuracy of the resulting models, and to find the probability of persons being at the age thresholds of legal relevance in a Thai population. A total of 1867 digital panoramic radiographs of Thai individuals aged between 8 and 23 years was selected to assess dental age. The mandibular third molar development was divided into nine stages. The stages were evaluated and each stage was transformed into a development score. Quadratic regression was employed to develop age prediction models. Our results show that males reached mandibular third molar root formation stages earlier than females. The models revealed a high correlation coefficient for both left and right mandibular third molar teeth in both sexes (R = 0.945 and 0.944 in males, R = 0.922 and 0.923 in females, respectively). Furthermore, the accuracy of the resulting models was tested in randomly selected 374 cases and showed low error values between the predicted dental age and the chronological age for both left and right mandibular third molar teeth in both sexes (−0.13 and −0.17 years in males, 0.01 and 0.03 years in females, respectively). In Thai samples, when the mandibular third molar teeth reached stage H, the probability of the person being over 18 years was 100 % in both sexes.

     

Sex estimation from knee breadth dimensions in a Finnish population

Sex estimation is an important part of osteological analysis of skeletons and forensic identification process. Traditionally cranial and pelvic traits are utilized for accurate sex estimation. However, post-cranial measurements have also been proven to accurately estimate sex especially from robust bones such as the femur. In this study, we investigated the potential of knee breadth dimensions in sex estimation in a Finnish population. To conduct this study we utilized a study sample (n = 1654) belonging to the Northern Finland Birth Cohort 1966. All individuals were 46 years of age at the time of the examination. Three knee breadth dimensions were measured from subjects' knee posteroanterior radiographs: femoral biepicondylar breadth (FBEB), mediolateral breadth of the femoral condyles (FCML), and mediolateral breadth of the tibial plateau (TPML). Sex estimation was performed using logistic regression. The study clearly demonstrated that all three measurements were different between males and females. Sectioning points for individual knee breadth measurements were 82.9 mm for FBEB, 76.6 mm for FCML and 75.4 mm for TPML. The classification rates ranged from 90.9% to 93.6%. The less commonly used measurements of FCML and TPML showed higher accuracy than FBEB in sex estimation. Our study confirmed that knee breadths can be successfully utilized to improve sex estimation in cases where the skeleton is only partially preserved and other major components of sex estimation are absent. We can also provide new standards for sex estimation from the knee joint in a Finnish population.     

Deep learning with convolutional neural network for estimation of the characterisation of coronary plaques: Validation using IB-IVUS

IntroductionDeep learning approaches have shown high diagnostic performance in image classifications, such as differentiation of malignant tumors and calcified coronary plaque. However, it is unknown whether deep learning is useful for characterizing coronary plaques without the presence of calcification using coronary computed tomography angiography (CCTA). The purpose of this study was to compare the diagnostic performance of deep learning with a convolutional neural network (CNN) with that of radiologists in the estimation of coronary plaques.MethodsWe retrospectively enrolled 178 patients (191 coronary plaques) who had undergone CCTA and integrated backscatter intravascular ultrasonography (IB-IVUS) studies. IB-IVUS diagnosed 81 fibrous and 110 fatty or fibro-fatty plaques. We manually captured vascular short-axis images of the coronary plaques as Portable Network Graphics (PNG) images (150 × 150 pixels). The display window level and width were 100 and 700 Hounsfield units (HU), respectively. The deep-learning system (CNN; GoogleNet Inception v3) was trained on 153 plaques; its performance was tested on 38 plaques. The area under the curve (AUC) obtained by receiver operating characteristic analysis of the deep learning system and by two board-certified radiologists was compared.ResultsWith the CNN, the AUC and the 95% confidence interval were 0.83 and 0.69–0.96, respectively; for radiologist 1 they were 0.61 and 0.42–0.80; for radiologist 2 they were 0.68 and 0.51–0.86, respectively. The AUC for CNN was significantly higher than for radiologists 1 (p = 0.04); for radiologist 2 it was not significantly different (p = 0.22).ConclusionDL-CNN performed comparably to radiologists for discrimination between fatty and fibro-fatty plaque on CCTA images.Implications for practiceThe diagnostic performance of the CNN and of two radiologists in the assessment of 191 ROIs on CT images of coronary plaques whose type corresponded with their IB-IVUS characterization was comparable.     

The Utility of a Convolutional Neural Network for Generating a Myelin Volume Index Map from Rapid Simultaneous Relaxometry Imaging

Purpose:A current algorithm to obtain a synthetic myelin volume fraction map (SyMVF) from rapid simultaneous relaxometry imaging (RSRI) has a potential problem, that it does not incorporate information from surrounding pixels. The purpose of this study was to develop a method that utilizes a convolutional neural network (CNN) to overcome this problem.Methods:RSRI and magnetization transfer images from 20 healthy volunteers were included. A CNN was trained to reconstruct RSRI-related metric maps into a myelin volume-related index (generated myelin volume index: GenMVI) map using the MVI map calculated from magnetization transfer images (MTMVI) as reference. The SyMVF and GenMVI maps were statistically compared by testing how well they correlated with the MTMVI map. The correlations were evaluated based on: (i) averaged values obtained from 164 atlas-based ROIs, and (ii) pixel-based comparison for ROIs defined in four different tissue types (cortical and subcortical gray matter, white matter, and whole brain).Results:For atlas-based ROIs, the overall correlation with the MTMVI map was higher for the GenMVI map than for the SyMVF map. In the pixel-based comparison, correlation with the MTMVI map was stronger for the GenMVI map than for the SyMVF map, and the difference in the distribution for the volunteers was significant (Wilcoxon sign-rank test, P < 0.001) in all tissue types.Conclusion:The proposed method is useful, as it can incorporate more specific information about local tissue properties than the existing method. However, clinical validation is necessary.     

Sex estimation from the clavicle using 3D reconstruction,discriminant analyses,and neural networks in an Eastern Turkish population

Sex estimation of skeletal remains is an important aspect of forensic anthropology. The clavicle is a bone with relatively high accuracy in sex determination. In this study, 7 clavicular parameters were obtained using the CT images and 3D reconstruction of 360 cases equally distributed as 180 males and 180 females. Sex determination was made using univariate, linear, and stepwise discriminant analyses, and multilayer perceptron neural networks. Maximum sex determination accuracy of 85.3% was achieved with univariate analysis, 89.4% with linear discriminant analysis, 90.0% with stepwise discriminant analysis, and 91.4% with multilayer perceptron neural networks. Significant changes were observed in the MLC, APMD-R and CDC parameters according to age, and these were determined to affect the accuracy of sex determination in different age groups. In forensic anthropological studies, more reliable results can be obtained by considering the confounding factors during sampling. Although high accuracy rates can be achieved with neural networks, the results should be approached with caution.     

An artificial intelligence deep learning model for identification of small bowel obstruction on plain abdominal radiographs

Objectives:Small bowel obstruction is a common surgical emergency which can lead to bowel necrosis, perforation and death. Plain abdominal X-rays are frequently used as a first-line test but the availability of immediate expert radiological review is variable. The aim was to investigate the feasibility of using a deep learning model for automated identification of small bowel obstruction.Methods:A total of 990 plain abdominal radiographs were collected, 445 with normal findings and 445 demonstrating small bowel obstruction. The images were labelled using the radiology reports, subsequent CT scans, surgical operation notes and enhanced radiological review. The data were used to develop a predictive model comprising an ensemble of five convolutional neural networks trained using transfer learning.Results:The performance of the model was excellent with an area under the receiver operator curve (AUC) of 0.961, corresponding to sensitivity and specificity of 91 and 93% respectively.Conclusion:Deep learning can be used to identify small bowel obstruction on plain radiographs with a high degree of accuracy. A system such as this could be used to alert clinicians to the presence of urgent findings with the potential for expedited clinical review and improved patient outcomes.Advances in knowledge:This paper describes a novel labelling method using composite clinical follow-up and demonstrates that ensemble models can be used effectively in medical imaging tasks. It also provides evidence that deep learning methods can be used to identify small bowel obstruction with high accuracy.     

Testing the nonlinear equations for dental age evaluation in a population of eastern China

ObjectiveThe study evaluated the validity of the nonlinear equations (Qingdao model) for dental age assessment in an eastern Chinese population.Materials and methodsWe studied 1073 digital panoramic radiographs of children aged 11–16 years from a Chinese Han population. Dental ages (DAs) were calculated using the Demirjian and the new model methods. Statistical significance was set at p < 0.05. For each method, differences between the chronological age (CA) and dental age were analyzed by paired t-tests and mean absolute error (MAE).ResultsThe discrepancies between CA and DA determined by Qingdao model were 0.18 and 0.30 years for males and females, respectively. While using Demirjian method, these differences were and 0.46 and 0.30. The Qingdao model’s MAEs between DA and CA were 1.23 and 0.90 years in males and females, respectively. As for the Demirjian method, MAEs were 1.43 and 0.86 years in males and females.ConclusionsThis study showed that the new nonlinear equations were more accurate than the traditional Demirjian method. Especially, the new nonlinear Qingdao model is more competitive in 11–14-year male groups and 15–16-year female groups. We recommend a combined Qingdao model and Demirjian method may reasonably reflect the CAs among children in the eastern Chinese population.     

Cranio-facial bones evaluation based on clinical CT data for sex determination in Northwest Indian population

The skull and pelvis have been the first choice of bones for determination of unknown human remains. The goal of the present study was to derive discriminant function equations by using clinical CT scan data of cranio-facial bones for sex determination in Northwest Indian population. This study was conducted at Department of Radiology, by collecting the retrospective data of CT scan of 217 samples. In the data, 106 were males and 111 were females in the age group between 20 and 80 years. The total number of parameters under investigation were 10. All the selected variables were sexually dimorphic and showed significant values. 91.7% of original grouped cases were correctly classified to their sex category. The TEM, rTEM and R were under the acceptable limits. The univariate, multivariate and stepwise discriminant function analysis recorded an accuracy of 88.9%, 91.7% and 93.6% respectively. Multivariate direct discriminant function analysis stepwise method yielded the highest level of accuracy in differentiating males and females. All the variables reflected statistically significant difference between males and females (p less than 0.001). The best single parameter with highest level of sexual dimorphic trait was length of cranial base. This study aims to provide sex assessment using clinical data of CT scan in Northwest Indian population by incorporating the BIOFB cranio-facial parameter. The morphometric measurements taken on CT scan images can be utilized by forensic experts in identification process.     

A Stacked Generalization of 3D Orthogonal Deep Learning Convolutional Neural Networks for Improved Detection of White Matter Hyperintensities in 3D FLAIR Images

BACKGROUND AND PURPOSE:Accurate and reliable detection of white matter hyperintensities and their volume quantification can provide valuable clinical information to assess neurologic disease progression. In this work, a stacked generalization ensemble of orthogonal 3D convolutional neural networks, StackGen-Net, is explored for improving automated detection of white matter hyperintensities in 3D T2-FLAIR images.MATERIALS AND METHODS:Individual convolutional neural networks in StackGen-Net were trained on 2.5D patches from orthogonal reformatting of 3D-FLAIR (n = 21) to yield white matter hyperintensity posteriors. A meta convolutional neural network was trained to learn the functional mapping from orthogonal white matter hyperintensity posteriors to the final white matter hyperintensity prediction. The impact of training data and architecture choices on white matter hyperintensity segmentation performance was systematically evaluated on a test cohort (n = 9). The segmentation performance of StackGen-Net was compared with state-of-the-art convolutional neural network techniques on an independent test cohort from the Alzheimer’s Disease Neuroimaging Initiative-3 (n = 20).RESULTS:StackGen-Net outperformed individual convolutional neural networks in the ensemble and their combination using averaging or majority voting. In a comparison with state-of-the-art white matter hyperintensity segmentation techniques, StackGen-Net achieved a significantly higher Dice score (0.76 [SD, 0.08], F1-lesion (0.74 [SD, 0.13]), and area under precision-recall curve (0.84 [SD, 0.09]), and the lowest absolute volume difference (13.3% [SD, 9.1%]). StackGen-Net performance in Dice scores (median = 0.74) did not significantly differ (P = .22) from interobserver (median = 0.73) variability between 2 experienced neuroradiologists. We found no significant difference (P = .15) in white matter hyperintensity lesion volumes from StackGen-Net predictions and ground truth annotations.CONCLUSIONS:A stacked generalization of convolutional neural networks, utilizing multiplanar lesion information using 2.5D spatial context, greatly improved the segmentation performance of StackGen-Net compared with traditional ensemble techniques and some state-of-the-art deep learning models for 3D-FLAIR.

White matter hyperintensities (WMHs) correspond to pathologic features of axonal degeneration, demyelination, and gliosis observed within cerebral white matter.¹ Clinically, the extent of WMHs in the brain has been associated with cognitive impairment, Alzheimer’s disease and vascular dementia, and increased risk of stroke.^2,3 The detection and quantification of WMH volumes to monitor lesion burden evolution and its correlation with clinical outcomes have been of interest in clinical research.^4,5 Although the extent of WMHs can be visually scored,⁶ the categoric nature of such scoring systems makes quantitative evaluation of disease progression difficult. Manually segmenting WMHs is tedious, prone to inter- and intraobserver variability, and is, in most cases, impractical. Thus, there is an increased interest in developing fast, accurate, and reliable computer-aided automated techniques for WMH segmentation.Convolutional neural network (CNN)-based approaches have been successful in several semantic segmentation tasks in medical imaging.⁷ Recent works have proposed using deep learning–based methods for segmenting WMHs using 2D-FLAIR images.^8-11 More recently, a WMH segmentation challenge¹² was also organized (http://wmh.isi.uu.nl/) to facilitate comparison of automated segmentation of WMHs of presumed vascular origin in 2D multislice T2-FLAIR images. Architectures that used an ensemble of separately trained CNNs showed promising results in this challenge, with 3 of the top 5 winners using ensemble-based techniques.¹²Conventional 2D-FLAIR images are typically acquired with thick slices (3–4 mm) and possible slice gaps. Partial volume effects from a thick slice are likely to affect the detection of smaller lesions, both in-plane and out-of-plane. 3D-FLAIR images, with isotropic resolution, have been shown to achieve higher resolution and contrast-to-noise ratio¹³ and have shown promising results in MS lesion detection using 3D CNNs.¹⁴ Additionally, the isotropic resolution enables viewing and evaluation of the images in multiple planes. This multiplanar reformatting of 3D-FLAIR without the use of interpolating kernels is only possible due to the isotropic nature of the acquisition. Network architectures that use information from the 3 orthogonal views have been explored in recent works for CNN-based segmentation of 3D MR imaging data.¹⁵ The use of data from multiple planes allows more spatial context during training without the computational burden associated with full 3D training.¹⁶ The use of 3 orthogonal views simultaneously mirrors how humans approach this segmentation task.Ensembles of CNNs have been shown to average away the variances in the solution and the choice of model- and configuration-specific behaviors of CNNs.¹⁷ Traditionally, the solutions from these separately trained CNNs are combined by averaging or using a majority consensus. In this work, we propose the use of a stacked generalization framework (StackGen-Net) for combining multiplanar lesion information from 3D CNN ensembles to improve the detection of WMH lesions in 3D-FLAIR. A stacked generalization¹⁸ framework learns to combine solutions from individual CNNs in the ensemble. We systematically evaluated the performance of this framework and compared it with traditional ensemble techniques, such as averaging or majority voting, and state-of-the-art deep learning techniques.