Improved Medical Image Segmentation Model Based on 3D U-Net

With the widespread application of deep learning in the field of computer vision,gradually allowing medical image technology to assist doctors in making diagnoses has great practical and research significance.Aiming at the shortcomings of the traditional U-Net model in 3D spatial information extraction,model over-fitting,and low degree of semantic information fusion,an improved medical image segmentation model has been used to achieve more accurate segmentation of medical images.In this model,we make full use of the residual network(ResNet)to solve the over-fitting problem.In order to process and aggregate data at different scales,the inception network is used instead of the traditional convolutional layer,and the dilated convolution is used to increase the receptive field.The conditional random field(CRF)can complete the contour refinement work.Compared with the traditional 3D U-Net network,the segmentation accuracy of the improved liver and tumor images increases by 2.89%and 7.66%,respectively.As a part of the image processing process,the method in this paper not only can be used for medical image segmentation,but also can lay the foundation for subsequent image 3D reconstruction work.

3-D Visualization of Medical Images with Arbitrary Sections

In this paper, with the general retrospect to the research on surface reconstruction and the marching cubes algorithm, we gave detailed description of an algorithm on the construction of object surfaces. The possible ambiguity problem in the original marching cubes algorithm was eliminated by its index mechanism. Some results on the MRI images were presented. Based on extracting and clipping contours from a set of medial slice images and setting the patch vertices values according to the gray images, this algorithm may be applied to form the arbitrary section images with three dimensional effects. It can also enhance the visual effect and interpretation of medical data.

Short‐term and long‐term memory self‐attention network for segmentation of tumours in 3D medical images

Tumour segmentation in medical images(especially 3D tumour segmentation)is highly challenging due to the possible similarity between tumours and adjacent tissues,occurrence of multiple tumours and variable tumour shapes and sizes.The popular deep learning‐based segmentation algorithms generally rely on the convolutional neural network(CNN)and Transformer.The former cannot extract the global image features effectively while the latter lacks the inductive bias and involves the complicated computation for 3D volume data.The existing hybrid CNN‐Transformer network can only provide the limited performance improvement or even poorer segmentation performance than the pure CNN.To address these issues,a short‐term and long‐term memory self‐attention network is proposed.Firstly,a distinctive self‐attention block uses the Transformer to explore the correlation among the region features at different levels extracted by the CNN.Then,the memory structure filters and combines the above information to exclude the similar regions and detect the multiple tumours.Finally,the multi‐layer reconstruction blocks will predict the tumour boundaries.Experimental results demonstrate that our method outperforms other methods in terms of subjective visual and quantitative evaluation.Compared with the most competitive method,the proposed method provides Dice(82.4%vs.76.6%)and Hausdorff distance 95%(HD95)(10.66 vs.11.54 mm)on the KiTS19 as well as Dice(80.2%vs.78.4%)and HD95(9.632 vs.12.17 mm)on the LiTS.

Research on a bifurcation location algorithm of a drainage tube based on 3D medical images

Based on patient computerized tomography data,we segmented a region containing an intracranial hematoma using the threshold method and reconstructed the 3D hematoma model.To improve the efficiency and accuracy of identifying puncture points,a point-cloud search arithmetic method for modified adaptive weighted particle swarm optimization is proposed and used for optimal external axis extraction.According to the characteristics of the multitube drainage tube and the clinical needs of puncture for intracranial hematoma removal,the proposed algorithm can provide an optimal route for a drainage tube for the hematoma,the precise position of the puncture point,and preoperative planning information,which have considerable instructional significance for clinicians.

CROSS-TRACK THREE APERTURES MILLIMETER WAVE SAR SIDE-LOOKING THREE-DIMENSIONAL IMAGING

The airborne cross-track three apertures MilliMeter Wave (MMW) Synthetic Aperture Radar (SAR) side-looking three-Dimensional (3D) imaging is investigated in this paper. Three apertures are distributed along the cross-track direction, and three virtual phase centers will be obtained through one-input and three-output. These three virtual phase centers form a sparse array which can be used to obtain the cross-track resolution. Because the cross-track array is short, the cross-track resolution is low. When the system works in side-looking mode, the cross-track resolution and height resolution will be coupling, and the low cross-track resolution will partly be transformed into the height uncertainty. The beam pattern of the real aperture is used as a weight to improve the Peak to SideLobe Ratio (PSLR) and Integrated SideLobe Ratio (ISLR) of the cross-track sparse array. In order to suppress the high cross-track sidelobes, a weighting preprocessing method is proposed. The 3D images of a point target and a simulation scene are achieved to verify the feasibility of the proposed method. And the imaging result of the real data obtained by the cross-track three-baseline MMW InSAR prototype is presented as a beneficial attempt.

A network lightweighting method for difficult segmentation of 3D medical images

Currently,deep learning is widely used in medical image segmentation and has achieved good results.However,3D medical image segmentation tasks with diverse lesion characters,blurred edges,and unstable positions require complex networks with a large number of parameters.It is computationally expensive and results in high requirements on equipment,making it hard to deploy the network in hospitals.In this work,we propose a method for network lightweighting and applied it to a 3D CNN based network.We experimented on a COVID-19 lesion segmentation dataset.Specifically,we use three cascaded one-dimensional convolutions to replace a 3D convolution,and integrate instance normalization with the previous layer of one-dimensional convolutions to accelerate network inference.In addition,we simplify test-time augmentation and deep supervision of the network.Experiments show that the lightweight network can reduce the prediction time of each sample and the memory usage by 50%and reduce the number of parameters by 60%compared with the original network.The training time of one epoch is also reduced by 50%with the segmentation accuracy dropped within the acceptable range.

3D Medical Image Segmentation Based on Rough Set Theory

This paper presents a method which uses multiple types of expert knowledge together in 3D medical image segmentation based on rough set theory. The focus of this paper is how to approximate a ROI(region of interest) when there are multiple types of expert knowledge. Based on rough set theory, the image can be split into three regions: positive regions; negative regions; boundary regions. With multiple knowledge we refine ROI as an intersection of all of the expected shapes with single knowledge. At last we show the results of implementing a rough 3D image segmentation and visualization system.

New Virtual Cutting Algorithms for 3D Surface Model Reconstructed from Medical Images

This paper proposes a practical algorithms of plane cutting, stereo clipping and arbitrary cutting for 3D surface model reconstructed from medical images. In plane cutting and stereo clipping algorithms, the 3D model is cut by plane or polyhedron. Lists of edge and vertex in every cut plane are established. From these lists the boundary contours are created and their relationship of embrace is ascertained. The region closed by the contours is triangulated using Delaunay triangulation algorithm. Arbitrary cutting operation creates cutting curve interactively. The cut model still maintains its correct topology structure. With these operations, tissues inside can be observed easily and it can aid doctors to diagnose. The methods can also be used in surgery planning of radiotherapy.

Monopulse instantaneous 3D imaging for wideband radar system

To avoid the complicated motion compensation in interferometric inverse synthetic aperture(InISAR)and achieve realtime three-dimensional(3 D)imaging,a novel approach for 3 D imaging of the target only using a single echo is presented.This method is based on an isolated scatterer model assumption,thus the scatterers in the beam can be extracted individually.The radial range of each scatterer is estimated by the maximal likelihood estimation.Then,the horizontal and vertical wave path difference is derived by using the phase comparison technology for each scatterer,respectively.Finally,by utilizing the relationship among the 3 D coordinates,the radial range,the horizontal and vertical wave path difference,the 3 D image of the target can be reconstructed.The reconstructed image is free from the limitation in InISAR that the image plane depends on the target's own motions and on its relative position with respect to the radar.Furthermore,a phase ambiguity resolution method is adopted to ensure the success of the 3 D imaging when phase ambiguity occurs.It can be noted that the proposed phase ambiguity resolution method only uses one antenna pair and does not require a priori knowledge,whereas the existing phase ambiguity methods may require two or more antenna pairs or a priori knowledge for phase unwarping.To evaluate the performance of the proposed method,the theoretical analyses on estimation accuracy are presented and the simulations in various scenarios are also carried out.

Three-Dimensional Model Reconstruction of Nonwovens from Multi-Focus Images

The three-dimensional(3D)model is of great significance to analyze the performance of nonwovens.However,the existing modelling methods could not reconstruct the 3D structure of nonwovens at low cost.A new method based on deep learning was proposed to reconstruct 3D models of nonwovens from multi-focus images.A convolutional neural network was trained to extract clear fibers from sequence images.Image processing algorithms were used to obtain the radius,the central axis,and depth information of fibers from the extraction results.Based on this information,3D models were built in 3D space.Furthermore,self-developed algorithms optimized the central axis and depth of fibers,which made fibers more realistic and continuous.The method with lower cost could reconstruct 3D models of nonwovens conveniently.

An image encryption scheme based on three-dimensional Brownian motion and chaotic system

At present, many chaos-based image encryption algorithms have proved to be unsafe, few encryption schemes permute the plain images as three-dimensional（3D） bit matrices, and thus bits cannot move to any position, the movement range of bits are limited, and based on them, in this paper we present a novel image encryption algorithm based on 3D Brownian motion and chaotic systems. The architecture of confusion and diffusion is adopted. Firstly, the plain image is converted into a 3D bit matrix and split into sub blocks. Secondly, block confusion based on 3D Brownian motion（BCB3DBM）is proposed to permute the position of the bits within the sub blocks, and the direction of particle movement is generated by logistic-tent system（LTS）. Furthermore, block confusion based on position sequence group（BCBPSG） is introduced, a four-order memristive chaotic system is utilized to give random chaotic sequences, and the chaotic sequences are sorted and a position sequence group is chosen based on the plain image, then the sub blocks are confused. The proposed confusion strategy can change the positions of the bits and modify their weights, and effectively improve the statistical performance of the algorithm. Finally, a pixel level confusion is employed to enhance the encryption effect. The initial values and parameters of chaotic systems are produced by the SHA 256 hash function of the plain image. Simulation results and security analyses illustrate that our algorithm has excellent encryption performance in terms of security and speed.

Airborne sparse flight array SAR 3D imaging based on compressed sensing in frequency domain

In airborne array synthetic aperture radar(SAR), the three-dimensional(3D) imaging performance and cross-track resolution depends on the length of the equivalent array. In this paper, Barker sequence criterion is used for sparse flight sampling of airborne array SAR, in order to obtain high cross-track resolution in as few times of flights as possible. Under each flight, the imaging algorithm of back projection(BP) and the data extraction method based on modified uniformly redundant arrays(MURAs) are utilized to obtain complex 3D image pairs. To solve the side-lobe noise in images, the interferometry between each image pair is implemented, and compressed sensing(CS) reconstruction is adopted in the frequency domain. Furthermore, to restore the geometrical relationship between each flight, the phase information corresponding to negative MURA is compensated on each single-pass image reconstructed by CS. Finally,by coherent accumulation of each complex image, the high resolution in cross-track direction is obtained. Simulations and experiments in X-band verify the availability.

DIRECT VOXEL-PROJECTION FOR VOLUMETRIC DATA RENDERING IN MEDICAL IMAGERY

The volumetric rendering of 3 D medical image data is very effective method for communication about radiological studies to clinicians. Algorithms that produce images with artifacts and inaccuracies are not clinically useful. This paper proposed a direct voxel projection algorithm to implement volumetric data rendering. Using this algorithm, arbitrary volume rotation, transparent and cutaway views are generated satisfactorily. Compared with the existing raytracing methods, it improves the projection image quality greatly. Some experimental results about real medical CT image data demonstrate the advantages and fidelity of the proposed algorithm.

A New Three-Dimensional(3D)Printing Prepress Algorithm for Simulation of Planned Surgery for Congenital Heart Disease

Background:Three-dimensional printing technology may become a key factor in transforming clinical practice and in significant improvement of treatment outcomes.The introduction of this technique into pediatric cardiac surgery will allow us to study features of the anatomy and spatial relations of a defect and to simulate the optimal surgical repair on a printed model in every individual case.Methods:We performed the prospective cohort study which included 29 children with congenital heart defects.The hearts and the great vessels were modeled and printed out.Measurements of the same cardiac areas were taken in the same planes and points at multislice computed tomography images(group 1)and on printed 3D models of the hearts(group 2).Pre-printing treatment of the multislice computed tomography data and 3D model preparation were performed according to a newly developed algorithm.Results:The measurements taken on the 3D-printed cardiac models and the tomographic images did not differ significantly,which allowed us to conclude that the models were highly accurate and informative.The new algorithm greatly simplifies and speeds up the preparation of a 3D model for printing,while maintaining high accuracy and level of detail.Conclusions:The 3D-printed models provide an accurate preoperative assessment of the anatomy of a defect in each case.The new algorithm has several important advantages over other available programs.They enable the development of customized preliminary plans for surgical repair of each specific complex congenital heart disease,predict possible issues,determine the optimal surgical tactics,and significantly improve surgical outcomes.

Reduced Imaging Time and Improved Image Quality of 3D Isotropic T2-Weighted Magnetic Resonance Imaging with Compressed Sensing for the Female Pelvis

This study is to compare three-dimensional(3D)isotropic T2-weighted magnetic resonance imaging(MRI)with compressed sensing-sampling perfection with application optimized contrast(CS-SPACE)and the conventional image(3D-SPACE)sequence in terms of image quality,estimated signal-to-noise ratio(SNR),relative contrast-to-noise ratio(CNR),and the lesions’conspicuous of the female pelvis.Thirty-six females(age:51,28-73)with cervical carcinoma(n=20),rectal carcinoma(n=7),or uterine fibroid(n=9)were included.Patients underwent magnetic resonance(MR)imaging at a 3T scanner with the sequences of 3D-SPACE,CS-SPACE,and twodimensional(2D)T2-weighted turbo-spin echo(TSE).Quantitative analyses of estimated SNR and relative CNR between tumors and other tissues,image quality,and tissue conspicuity were performed.Two radiologists assessed the difference in diagnostic findings for carcinoma.Quantitative values and qualitative scores were analyzed,respectively.The estimated SNR and the relative CNR of tumor-to-muscle obturator internus,tumor-to-myometrium,and myometrium-to-muscle obturator internus was comparable between 3D-SPACE and CS-SPACE.The overall image quality and the conspicuity of the lesion scores of the CS-SPACE were higher than that of the 3D-SPACE(P<0.01).The CS-SPACE sequence offers shorter scan time,fewer artifacts,and comparable SNR and CNR to conventional 3D-SPACE,and has the potential to improve the performance of T2-weighted images.

Designing a High-Performance Deep Learning Theoretical Model for Biomedical Image Segmentation by Using Key Elements of the Latest U-Net-Based Architectures

Deep learning (DL) has experienced an exponential development in recent years, with major impact in many medical fields, especially in the field of medical image and, respectively, as a specific task, in the segmentation of the medical image. We aim to create a computer assisted diagnostic method, optimized by the use of deep learning (DL) and validated by a randomized controlled clinical trial, is a highly automated tool for diagnosing and staging precancerous and cervical cancer and thyroid cancers. We aim to design a high-performance deep learning model, combined from convolutional neural network (U-Net)-based architectures, for segmentation of the medical image that is independent of the type of organs/tissues, dimensions or type of image (2D/3D) and to validate the DL model in a randomized, controlled clinical trial. We used as a methodology primarily the analysis of U-Net-based architectures to identify the key elements that we considered important in the design and optimization of the combined DL model, from the U-Net-based architectures, imagined by us. Secondly, we will validate the performance of the DL model through a randomized controlled clinical trial. The DL model designed by us will be a highly automated tool for diagnosing and staging precancers and cervical cancer and thyroid cancers. The combined model we designed takes into account the key features of each of the architectures Overcomplete Convolutional Network Kite-Net (Kite-Net), Attention gate mechanism is an improvement added on convolutional network architecture for fast and precise segmentation of images (Attention U-Net), Harmony Densely Connected Network-Medical image Segmentation (HarDNet-MSEG). In this regard, we will create a comprehensive computer assisted diagnostic methodology validated by a randomized controlled clinical trial. The model will be a highly automated tool for diagnosing and staging precancers and cervical cancer and thyroid cancers. This would help drastically minimize the time and effort that specialists put into analyzing medical images, help to achieve a better therapeutic plan, and can provide a “second opinion” of computer assisted diagnosis.

Visualization of Three-dimensional Human Data Based on CT Image

Three-dimensional medical image visualization becomes an essential part for medical field, including computer aided diagnosis, surgery planning and simulation, artificial limb surgery, radiotherapy planning, and teaching etc. In this paper, marching cubes algorithm is adopted to reconstruct the 3-D images for the CT image sequence in DICOM format under theVC＋＋6.0 and the visual package VTK platform. The relatively simple interactive operations such as rotation and transfer can be realized on the platform. Moreover, the normal vector and interior point are calculated to form the virtual clipping plane, which is then used to incise the 3-D object. Information of the virtual slice can be obtained, in the mean while the virtual slice images are displayed on the screen. The technique can realize the real time interaction extraction of virtual slice on 3-D CT image. The cuboids structured can be zoomed, moved and eircumrotated by operating mouse to incise the 3-D reconstruction object. Real time interaction can be realized by clipping the reconstruction object. The coordinates can be acquired by the mouse clicking in the 3D space, to realize the point mouse pick-up as well angle and distance interactive measurement. We can get quantitative information about 3-D images through measurement.

Design and implementation of GM- APD array readout circuit for infrared imaging

Based on an avalanche photodiode（ APD） detecting array working in Geiger mode（ GM-APD）, a high-performance infrared sensor readout integrated circuit（ ROIC） used for infrared 3D（ three-dimensional） imaging is proposed. The system mainly consists of three functional modules, including active quenching circuit（ AQC）, time-to-digital converter（ TDC） circuit and other timing controller circuit. Each AQC and TDC circuit together constitutes the pixel circuit. Under the cooperation with other modules, the current signal generated by the GM-APD sensor is detected by the AQC, and the photon time-of-flight（ TOF） is measured and converted to a digital signal output to achieve a better noise suppression and a higher detection sensitivity by the TDC. The ROIC circuit is fabricated by the CSMC 0. 5 μm standard CMOS technology. The array size is 8 × 8, and the center distance of two adjacent cells is 100μm. The measurement results of the chip showthat the performance of the circuit is good, and the chip can achieve 1 ns time resolution with a 250 MHz reference clock, and the circuit can be used in the array structure of the infrared detection system or focal plane array（ FPA）.

Super-resolution amplification of bitmap images based on 3D modeling theory

As a general format of the image,bitmap(BMP)image has wide applications,and consequently it is an important part of image processing.By segmenting the bitmap and combining the three-dimesional(3D)model of the discrete algorithm with the scanning line compensation algorithm,a mathematical model is built.According to the topological relations between several control points on the model surface,the surface of the model is discretized,and a planar triangle sequence is used to describe 3D objects.Finally,the bitmap is enlarged by combining the borrowing compensation based on 3D modeling principle of discrete algorithm with the scanning line compensation algorithm of binary lattice image,thus getting a relatively clear enlarged BMP image.

磁共振3D神经成像对腰骶丛显示的对比研究

目的探讨1.5T磁共振三维快速自旋回波成像(3D sampling perfection with application optimized contrasts by using different flip angle evolutions,3D-SPACE)与多回波融合成像(multiple echo data image combination,3D-MEDIC)序列对正常人腰骶丛神经的显示。方法 31例无症状正常志愿者行MRI检查,包括常规腰椎MRI、3D-SPACE及MEDIC序列扫描,原始图像传入后处理工作站行多平面图像重组。结合原始及重组图像,比较3D-SPACE及MEDIC序列图像腰骶丛神经根的信噪比及对神经的显示评分。结果 3D-MEDIC序列显示神经根的信噪比高于3D-SPACE序列(分别为70.15±24.03及28.78±7.12,P=0.000)。3D-MEDIC序列对腰骶丛的显示评分高于3D-SPACE序列(P=0.000)。结论1.5T磁共振腰骶丛神经显像中,3D-MEDIC序列优于3D-SPACE序列,可清晰显示神经根走行,是常规腰椎MRI的重要补充。