Classification of white blood cells using weighted optimized deformable convolutional neural networks

BACKGROUND

Machine learning (ML) algorithms have been widely used in the classification of white blood cells (WBCs). However, the performance of ML algorithms still needs to be addressed for being short of gold standard data sets, and even the implementation of the proposed algorithms.

METHODS

In this study, the method of two-module weighted optimized deformable convolutional neural networks (TWO-DCNN) was proposed for WBC classification. Our algorithm is characterized as two-module transfer learning and deformable convolutional (DC) layers for the betterment of robustness. To validate the performance, our method was compared with classical MLs of VGG16, VGG19, Inception-V3, ResNet-50, support vector machine (SVM), multilayer perceptron (MLP), decision tree (DT) and random forest (RF) on our undisclosed WBC data set and public BCCD data set.

RESULTS

TWO-DCNN achieved the best performance with the precisions (PREs) of 95.7%, 94.5% and 91.6%, recalls (RECs) of 95.7%, 94.5% and 91.6%, F1-scores (F1s) of 95.7%, 94.5% and 91.6%, area under curves (AUCs) of 0.98, 0.97 and 0.95 for low-resolution and noisy undisclosed data sets, BCCD data set, respectively.

CONCLUSIONS

With accurate feature extraction and optimized network weights, the proposed TWO-DCNN showed the best performance in WBC classification for low-resolution and noisy data sets. It could be used as an alternative method for clinical applications.

Diffusion tensor imaging fiber tracking with reliable tracking orientation and flexible step size

We propose a method of reliable tracking orientation and flexible step size fiber tracking. A new directional strategy was defined to select one optimal tracking orientation from each directional set, which was based on the single-tensor model and the two-tensor model. The directional set of planar voxels contained three tracking directions: two from the two-tensor model and one from the single-tensor model. The directional set of linear voxels contained only one principal vector. In addition, a flexible step size, rather than fixable step sizes, was implemented to improve the accuracy of fiber tracking. We used two sets of human data to assess the performance of our method; one was from a healthy volunteer and the other from a patient with low-grade glioma. Results verified that our method was superior to the single-tensor Fiber Assignment by Continuous Tracking and the two-tensor eXtended Streamline Tractography for showing detailed images of fiber bundles.

Automatic recognition of cephalometric landmarks via multi-scale sampling strategy

The identification of head landmarks in cephalometric analysis significantly contributes in the anatomical localization of maxillofacial tissues for orthodontic and orthognathic surgery. However, the existing methods face the limitations of low accuracy and cumbersome identification process. In this pursuit, the present study proposed an automatic target recognition algorithm called Multi-Scale YOLOV3 (MS-YOLOV3) for the detection of cephalometric landmarks. It was characterized by multi-scale sampling strategies for shallow and deep features at varied resolutions, and especially contained the module of spatial pyramid pooling (SPP) for highest resolution. The proposed method was quantitatively and qualitatively compared with the classical YOLOV3 algorithm on the two data sets of public lateral cephalograms, undisclosed anterior-posterior (AP) cephalograms, respectively, for evaluating the performance. The proposed MS-YOLOV3 algorithm showed better robustness with successful detection rates (SDR) of 80.84% within 2 mm, 93.75% within 3 mm, and 98.14% within 4 mm for lateral cephalograms, and 85.75% within 2 mm, 92.87% within 3 mm, and 96.66% within 4 mm for AP cephalograms, respectively. It was concluded that the proposed model could be robustly used to label the cephalometric landmarks on both lateral and AP cephalograms for the clinical application in orthodontic and orthognathic surgery.

Revealing the correlations between brain cortical characteristics and susceptibility genes for Alzheimer disease: a cross-sectional study

Background

Alzheimer disease (AD) is a progressive neurodegenerative disease closely related to genes and characterized by the atrophy of the cerebral cortex. Correlations between imaging phenotypes and the susceptibility genes for AD, as demonstrated in the findings of genome-wide association studies (GWASs), still need to be addressed due to the complicated structure of the human cortex.

Methods

In our study, an improved GWAS method, whole cortex characteristics GWAS (WCC-GWAS), was proposed. The WCC-GWAS uses multiple cortex characteristics of gray-matter volume (GMV), cortical thickness (CT), cortical surface area (CSA), and local gyrification index (LGI). A cohort of 496 participants was enrolled and divided into 4 groups: normal control (NC; n=122), early mild cognitive impairment (EMCI; n=196), late mild cognitive impairment (LMCI; n=62), and AD (n=116). Based on the Desikan-Killiany atlas, the brain was parcellated into 68 brain regions, and the WCC of each brain region was individually calculated. Four cortex characteristics of GMV, CT, CSA, and LGI across the 4 groups optimized with multiple comparisons and the ReliefF algorithm were taken as magnetic resonance imaging (MRI) brain phenotypes. Under the model of multiple linear additive genetic regression, the correlations between the MRI brain phenotypes and single-nucleotide polymorphisms (SNPs) were deduced.

Results

The findings identified 2 prominent correlations. First, rs7309929 of neuron navigator 3 (NAV3) located on chromosome 12 correlated with the decreased GMV for the left middle temporal gyrus (P=0.0074). Second, rs11250992 of long intergenic non-protein-coding RNA 700 (LINC00700) located on chromosome 10 correlated with the decreased CT for the left supramarginal gyrus (P=0.0019).

Conclusions

The findings suggested that the correlations between phenotypes and genotypes could be effectively evaluated. The strategy of extracting MRI phenotypes as endophenotypes provided valuable indications in AD GWAS.