An optimised YOLOv4 deep learning model for efficient malarial cell detection in thin blood smear images

BACKGROUND

Malaria is a serious public health concern worldwide. Early and accurate diagnosis is essential for controlling the disease's spread and avoiding severe health complications. Manual examination of blood smear samples by skilled technicians is a time-consuming aspect of the conventional malaria diagnosis toolbox. Malaria persists in many parts of the world, emphasising the urgent need for sophisticated and automated diagnostic instruments to expedite the identification of infected cells, thereby facilitating timely treatment and reducing the risk of disease transmission. This study aims to introduce a more lightweight and quicker model-but with improved accuracy-for diagnosing malaria using a YOLOv4 (You Only Look Once v. 4) deep learning object detector.

METHODS

The YOLOv4 model is modified using direct layer pruning and backbone replacement. The primary objective of layer pruning is the removal and individual analysis of residual blocks within the C3, C4 and C5 (C3-C5) Res-block bodies of the backbone architecture's C3-C5 Res-block bodies. The CSP-DarkNet53 backbone is simultaneously replaced for enhanced feature extraction with a shallower ResNet50 network. The performance metrics of the models are compared and analysed.

RESULTS

The modified models outperform the original YOLOv4 model. The YOLOv4-RC3_4 model with residual blocks pruned from the C3 and C4 Res-block body achieves the highest mean accuracy precision (mAP) of 90.70%. This mAP is > 9% higher than that of the original model, saving approximately 22% of the billion floating point operations (B-FLOPS) and 23 MB in size. The findings indicate that the YOLOv4-RC3_4 model also performs better, with an increase of 9.27% in detecting the infected cells upon pruning the redundant layers from the C3 Res-block bodies of the CSP-DarkeNet53 backbone.

CONCLUSIONS

The results of this study highlight the use of the YOLOv4 model for detecting infected red blood cells. Pruning the residual blocks from the Res-block bodies helps to determine which Res-block bodies contribute the most and least, respectively, to the model's performance. Our method has the potential to revolutionise malaria diagnosis and pave the way for novel deep learning-based bioinformatics solutions. Developing an effective and automated process for diagnosing malaria will considerably contribute to global efforts to combat this debilitating disease. We have shown that removing undesirable residual blocks can reduce the size of the model and its computational complexity without compromising its precision.

TL-MSE2-Net: Transfer learning based nested model for cerebrovascular segmentation with aneurysms

Cerebrovascular (i.e., cerebral vessel) segmentation is essential for diagnosing and treating brain diseases. Convolutional neural network models, such as U-Net, are commonly used for this purpose. Unfortunately, such models may not be entirely satisfactory in dealing with cerebrovascular segmentation with tumors due to the following issues: (1) Relatively small number of clinical datasets from patients obtained through different modalities such as computed tomography (CT) and magnetic resonance imaging (MRI), leading to inadequate training and lack of transferability in the modeling; (2) Insufficient feature extraction caused by less attention to both convolution sizes and cerebral vessel edges. Inspired by the existence of similar features on cerebral vessels between normal subjects and patients, we propose a transfer learning strategy based on a pre-trained nested model called TL-MSE2-Net. This model uses one of the publicly available datasets for cerebrovascular segmentation with aneurysms. To address issue (1), our transfer learning strategy leverages a pre-trained model that uses a large number of datasets from normal subjects, providing a potential solution to the lack of sufficient clinical datasets. To tackle issue (2), we structure the pre-trained model based on 3D U-Net, comprising three blocks: ResMul, DeRes, and REAM. The ResMul and DeRes blocks enhance feature extraction by utilizing multiple convolution sizes to capture multiscale features, and the REAM block increases the weight of the voxels on the edges of the given 3D volume. We evaluated the proposed model on one small private clinical dataset and two publicly available datasets. The experimental results demonstrated that our MSE2-Net framework achieved an average Dice score of 70.81 % and 89.08 % on the two publicly available datasets, outperforming other state-of-the-art methods. Ablation studies were also conducted to validate the effectiveness of each block. The proposed TL-MSE2-Net yielded better results than MSE2-Net on a small private clinical dataset, with increases of 5.52 %, 3.37 %, 6.71 %, and 0.85 % for the Dice score, sensitivity, Jaccard index, and precision, respectively.

PrUb-EL: A hybrid framework based on deep learning for identifying ubiquitination sites in Arabidopsis thaliana using ensemble learning strategy

Identification of ubiquitination sites is central to many biological experiments. Ubiquitination is a kind of post-translational protein modification (PTM). It is a key mechanism for increasing protein diversity and plays a vital role in regulating cell function. In recent years, many models have been developed to predict ubiquitination sites in humans, mice and yeast. However, few studies have predicted ubiquitination sites in Arabidopsis thaliana. In view of this, a deep network model named PrUb-EL is proposed to predict ubiquitination sites in Arabidopsis thaliana. Firstly, six features based on the protein sequence are extracted with amino acid index database (AAindex), dipeptide deviates from the expected mean (DDE), dipeptide composition (DPC), blocks substitution matrix (BLOSUM62), enhanced amino acid composition (EAAC) and binary encoding. Secondly, the synthetic minority over-sampling technique (SMOTE) is utilized to process the imbalanced data set. Then a new classifier named DG is presented, which includes Dense block, Residual block and Gated recurrent unit (GRU) block. Finally, each of six feature extraction methods is integrated into the DG model, and the ensemble learning strategy is used to gain the final prediction result. Experimental results show that PrUb-EL has good predictive ability with the accuracy (ACC) and area under the ROC curve (auROC) values of 91.00% and 97.70% using 5-fold cross-validation, respectively. Note that the values of ACC and auROC are 88.58% and 96.09% in the independent test, respectively. Compared with previous studies, our model has significantly improved performance thus it is an excellent method for identifying ubiquitination sites in Arabidopsis thaliana. The datasets and code used for the article are available at https://github.com/Tom-Wangy/PreUb-EL.git.

[Research on three-dimensional skull repair by combining residual and informer attention]

Cranial defects may result from clinical brain tumor surgery or accidental trauma. The defect skulls require hand-designed skull implants to repair. The edge of the skull implant needs to be accurately matched to the boundary of the skull wound with various defects. For the manual design of cranial implants, it is time-consuming and technically demanding, and the accuracy is low. Therefore, an informer residual attention U-Net (IRA-Unet) for the automatic design of three-dimensional (3D) skull implants was proposed in this paper. Informer was applied from the field of natural language processing to the field of computer vision for attention extraction. Informer attention can extract attention and make the model focus more on the location of the skull defect. Informer attention can also reduce the computation and parameter count from N 2 to log( N). Furthermore,the informer residual attention is constructed. The informer attention and the residual are combined and placed in the position of the model close to the output layer. Thus, the model can select and synthesize the global receptive field and local information to improve the model accuracy and speed up the model convergence. In this paper, the open data set of the AutoImplant 2020 was used for training and testing, and the effects of direct and indirect acquisition of skull implants on the results were compared and analyzed in the experimental part. The experimental results show that the performance of the model is robust on the test set of 110 cases fromAutoImplant 2020. The Dice coefficient and Hausdorff distance are 0.940 4 and 3.686 6, respectively. The proposed model reduces the resources required to run the model while maintaining the accuracy of the cranial implant shape, and effectively assists the surgeon in automating the design of efficient cranial repair, thereby improving the quality of the patient's postoperative recovery.

[A computed tomography image segmentation algorithm for improving the diagnostic accuracy of rectal cancer based on U-net and residual block]

As an important basis for lesion determination and diagnosis, medical image segmentation has become one of the most important and hot research fields in the biomedical field, among which medical image segmentation algorithms based on full convolutional neural network and U-Net neural network have attracted more and more attention by researchers. At present, there are few reports on the application of medical image segmentation algorithms in the diagnosis of rectal cancer, and the accuracy of the segmentation results of rectal cancer is not high. In this paper, a convolutional network model of encoding and decoding combined with image clipping and pre-processing is proposed. On the basis of U-Net, this model replaced the traditional convolution block with the residual block, which effectively avoided the problem of gradient disappearance. In addition, the image enlargement method is also used to improve the generalization ability of the model. The test results on the data set provided by the "Teddy Cup" Data Mining Challenge showed that the residual block-based improved U-Net model proposed in this paper, combined with image clipping and preprocessing, could greatly improve the segmentation accuracy of rectal cancer, and the Dice coefficient obtained reached 0.97 on the verification set.

Deep learning approach for automatic segmentation of ulna and radius in dual-energy X-ray imaging

BACKGROUND

Segmentation of the ulna and radius is a crucial step for the measurement of bone mineral density (BMD) in dual-energy X-ray imaging in patients suspected of having osteoporosis.

PURPOSE

This work aimed to propose a deep learning approach for the accurate automatic segmentation of the ulna and radius in dual-energy X-ray imaging.

METHODS AND MATERIALS

We developed a deep learning model with residual block (Resblock) for the segmentation of the ulna and radius. Three hundred and sixty subjects were included in the study, and five-fold cross-validation was used to evaluate the performance of the proposed network. The Dice coefficient and Jaccard index were calculated to evaluate the results of segmentation in this study.

RESULTS

The proposed network model had a better segmentation performance than the previous deep learning-based methods with respect to the automatic segmentation of the ulna and radius. The evaluation results suggested that the average Dice coefficients of the ulna and radius were 0.9835 and 0.9874, with average Jaccard indexes of 0.9680 and 0.9751, respectively.

CONCLUSION

The deep learning-based method developed in this study improved the segmentation performance of the ulna and radius in dual-energy X-ray imaging.

A residual network with attention module for hyperspectral information of recognition to trace the origin of rice

In this work, a neural network framework for hyperspectral information recognition was proposed, combined with residual block and convolutional block attention module (CBAM) to enhance the detection performance of hyperspectral for tracing the rice quality. Firstly, the hyperspectral image system was used to obtain the hyperspectral information of the rice. Secondly, due to the small data set, the structure of the residual network was designed based on the characteristics of the hyperspectral information to prevent overfitting the model. Finally, the CBAM was introduced to calculate the channel and spatial attention to redistribute the weight parameter and enhance the classification performance of the model. The results showed that our (Res-CBAM) model had better classification performance than other classification methods. The classification accuracy of the rice was 96.33%. This study provided a strategy to enhance the detection performance of hyperspectral, and an intelligent technology to trace the rice quality.

Rigid and non-rigid motion artifact reduction in X-ray CT using attention module

Motion artifacts are a major factor that can degrade the diagnostic performance of computed tomography (CT) images. In particular, the motion artifacts become considerably more severe when an imaging system requires a long scan time such as in dental CT or cone-beam CT (CBCT) applications, where patients generate rigid and non-rigid motions. To address this problem, we proposed a new real-time technique for motion artifacts reduction that utilizes a deep residual network with an attention module. Our attention module was designed to increase the model capacity by amplifying or attenuating the residual features according to their importance. We trained and evaluated the network by creating four benchmark datasets with rigid motions or with both rigid and non-rigid motions under a step-and-shoot fan-beam CT (FBCT) or a CBCT. Each dataset provided a set of motion-corrupted CT images and their ground-truth CT image pairs. The strong modeling power of the proposed network model allowed us to successfully handle motion artifacts from the two CT systems under various motion scenarios in real-time. As a result, the proposed model demonstrated clear performance benefits. In addition, we compared our model with Wasserstein generative adversarial network (WGAN)-based models and a deep residual network (DRN)-based model, which are one of the most powerful techniques for CT denoising and natural RGB image deblurring, respectively. Based on the extensive analysis and comparisons using four benchmark datasets, we confirmed that our model outperformed the aforementioned competitors. Our benchmark datasets and implementation code are available at https://github.com/youngjun-ko/ct_mar_attention.

econvRBP: Improved ensemble convolutional neural networks for RNA binding protein prediction directly from sequence

RNA binding proteins (RBPs) determine RNA process from synthesis to decay, which play a key role in RNA transport, translation and degradation. Therefore, exploring RBPs' function from the amino acid sequence using computational methods has become one of the momentous topics in genome annotation. However, there still have some challenges: (1) shallow feature: Although the sequence determines structure is self-evident, it is difficult to analyze the essential features from simple sequence. (2) Poorly understand: feature-based prediction methods mainly emphasize feature extraction, while in-depth understanding of protein mysteries limits the application of feature engineering. (3) Feature fusion: multi-feature fusion is often used, but the features are not well integrated. In view of these challenges, we propose a novel ensemble convolutional neural network (econvRBP) to predict RBPs. In order to capture the local and global features of RNA binding proteins simultaneously, first of all, One Hot and Conjoint Triad encoding methods are used to transform amino acid sequence into local and global features, respectively. After that the local and global features are combined for further high-level feature extraction using convolutional neural networks. Some experiments are constructed to evaluate our method with 10-fold cross validation and the results show that it has achieved the best performance among all the predictors so far. We correctly predicted 99% of 2875 RBPs and 99% of 6782 non-RBPs with accuracy of 0.99. In addition, the datasets provided by RBPPred are also used to validate our models with an accuracy of 0.87. These results indicate that the econvRBP is the most excellent method at present, and will provide reliable guidance for the detection of RBPs. econvRBP is available at http://47.100.203.218:3389/home.html/.

Multi-view secondary input collaborative deep learning for lung nodule 3D segmentation

BACKGROUND

Convolutional neural networks (CNNs) have been extensively applied to two-dimensional (2D) medical image segmentation, yielding excellent performance. However, their application to three-dimensional (3D) nodule segmentation remains a challenge.

METHODS

In this study, we propose a multi-view secondary input residual (MV-SIR) convolutional neural network model for 3D lung nodule segmentation using the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset of chest computed tomography (CT) images. Lung nodule cubes are prepared from the sample CT images. Further, from the axial, coronal, and sagittal perspectives, multi-view patches are generated with randomly selected voxels in the lung nodule cubes as centers. Our model consists of six submodels, which enable learning of 3D lung nodules sliced into three views of features; each submodel extracts voxel heterogeneity and shape heterogeneity features. We convert the segmentation of 3D lung nodules into voxel classification by inputting the multi-view patches into the model and determine whether the voxel points belong to the nodule. The structure of the secondary input residual submodel comprises a residual block followed by a secondary input module. We integrate the six submodels to classify whether voxel points belong to nodules, and then reconstruct the segmentation image.

RESULTS

The results of tests conducted using our model and comparison with other existing CNN models indicate that the MV-SIR model achieves excellent results in the 3D segmentation of pulmonary nodules, with a Dice coefficient of 0.926 and an average surface distance of 0.072.

CONCLUSION

our MV-SIR model can accurately perform 3D segmentation of lung nodules with the same segmentation accuracy as the U-net model.