Interpolation linearization predistortion technology for FMCW LiDAR.

Frequency-modulated continuous-wave (FMCW) laser ranging technology is an important development direction of light detection and ranging (LiDAR) for the future. It has the advantages of high ranging accuracy, high resolution, wide range, and no ranging blind zone. A distributed feedback laser can be used as a high-quality light source in FMCW laser ranging systems because of its wide frequency modulation range, simple frequency modulation mode, and small package. Aiming at the nonlinear problem of the laser in the frequency modulation process, we present a novel, to our knowledge, predistortion algorithm based on interpolation linear fitting to enhance the linearity of the FMCW laser for LiDAR systems. The sweeping frequency curve of the laser is obtained using the Hilbert transform, and then the sweeping frequency curve is segmented and linearly fitted to calculate the interpolated driving current signals corresponding to linear frequency changes. Using this method, we achieved a nonlinearity error lower than 1e-7 for the swept-frequency signal and demonstrated that the ranging error is less than ±5c m at a distance of 100 m in the FMCW system. In addition, we also demonstrated a 3D static object point cloud with high imaging quality. Compared with the iterative predistortion algorithm based on the function fitting, this method avoids fitting errors at the inflection points of the triangular swept-frequency signal and the complexity of multiple iterative calculations. It enables rapid generation of pre-distorted swept-frequency signals, making it particularly suitable for real-time applications of automotive LiDAR systems.

Patient-Reported Outcomes and Complication Profiles of Implant-Based Breast Reconstruction in Patients With Postmastectomy Radiation Therapy.

OBJECTIVE: This study aims to investigate the patient-reported outcomes (PROs) and complications of distinct implant-based breast reconstruction modality for patients with postmastectomy radiation therapy (PMRT). METHODS: A retrospective review was conducted on breast cancer patients with stage II-III disease who performed implant-based breast reconstruction following with PMRT between September 2016 and April 2022. The patients were categorized into two matched groups: (1) patients receiving prepectoral breast reconstruction (PBR) or (2) subpectoral breast reconstruction (SBR) followed by PMRT. Following reconstruction, the patients were further compared for PMRT with the tissue expander (PMRT-TE) versus PMRT with permanent implant (PMRT-PI). PROs were measured with BREAST-Q questionnaire. Early and late complications were recorded and analyzed. RESULTS: A total of 55 eligible patients were recruited. Patients who underwent PBR reported significantly higher satisfaction with breasts scores (P = 0.003) compared with the SBR group. The PMRT-TE group had higher satisfaction with breasts (P = 0.001) but lower physical well-being (P = 0.029) scores compared with PMRT-PI group. Moreover, patients in SBR cohort had a higher risk of capsular contracture (Baker grade III or IV) (20.5% vs 6.3%) and implant dislocation (48.7% vs 12.5%) than patients in PBR cohort. Patients in PMRT-PI group had a slightly higher rate of capsular contracture (Baker grade III or IV) than PMRT-TE group (20.8% vs 12.9%). CONCLUSIONS: PBR was associated with lower rates of late complications, especially for implant dislocation, and higher satisfaction with breasts scores compared to SBR. In addition, compared to PMRT-TE with PMRT-PI, patients in PMRT-TE cohort reported superior PROs of satisfaction with breasts.

Few-Shot Object Detection in Remote Sensing Images via Data Clearing and Stationary Meta-Learning.

Nowadays, the focus on few-shot object detection (FSOD) is fueled by limited remote sensing data availability. In view of various challenges posed by remote sensing images (RSIs) and FSOD, we propose a meta-learning-based Balanced Few-Shot Object Detector (B-FSDet), built upon YOLOv9 (GELAN-C version). Firstly, addressing the problem of incompletely annotated objects that potentially breaks the balance of the few-shot principle, we propose a straightforward yet efficient data clearing strategy, which ensures balanced input of each category. Additionally, considering the significant variance fluctuations in output feature vectors from the support set that lead to reduced effectiveness in accurately representing object information for each class, we propose a stationary feature extraction module and corresponding stationary and fast prediction method, forming a stationary meta-learning mode. In the end, in consideration of the issue of minimal inter-class differences in RSIs, we propose inter-class discrimination support loss based on the stationary meta-learning mode to ensure the information provided for each class from the support set is balanced and easier to distinguish. Our proposed detector's performance is evaluated on the DIOR and NWPU VHR-10.v2 datasets, and comparative analysis with state-of-the-art detectors reveals promising performance.

Underwater image enhancement via red channel maximum attenuation prior and multi-scale detail fusion.

The underwater environment poses great challenges, which have a negative impact on the capture and processing of underwater images. However, currently underwater imaging systems cannot adapt to various underwater environments to guarantee image quality. To address this problem, this paper designs an efficient underwater image enhancement approach that gradually adjusts colors, increases contrast, and enhances details. Based on the red channel maximum attenuation prior, we initially adjust the blue and green channels and correct the red channel from the blue and green channels. Subsequently, the maximum and minimum brightness blocks are estimated in multiple channels to globally stretch the image, which also includes our improved guided noise reduction filtering. Finally, in order to amplify local details without affecting the naturalness of the results, we use a pyramid fusion model to fuse local details extracted from two methods, taking into account the detail restoration effect of the optical model. The enhanced underwater image through our method has rich colors without distortion, effectively improved contrast and details. The objective and subjective evaluations indicate that our approach surpasses the state-of-the-art methods currently. Furthermore, our approach is versatile and can be applied to diverse underwater scenes, which facilitates subsequent applications.

TREM2 expression on the microglia resolved lead exposure-induced neuroinflammation by promoting anti-inflammatory activities.

Neurotoxicity caused by environmental lead (Pb) pollution is a worldwide public health concern, and developing a therapeutic strategy against Pb-induced neurotoxicity is an important area in the current research. Our prior research has demonstrated the significant involvement of microglia-mediated inflammatory responses in the manifestation of Pb-induced neurotoxicity. Additionally, the suppression of proinflammatory mediator activity significantly mitigated the toxic effects associated with Pb exposure. Recent studies have highlighted the critical role of the triggering receptor expressed on myeloid cells 2 (TREM2) in the pathogenesis of neurodegenerative disorders. TREM2 exerted protective effects on inflammation, but whether TREM2 is involved in Pb-induced neuroinflammation is poorly understood. In the present study, cell culture experiments and animal models were designed to investigate the role of TREM2 in Pb's neuroinflammation. We examined the impact of pro- and anti-inflammatory cytokines involved in Pb-induced neuroinflammation. Flow cytometry and microscopy techniques were applied to detect microglia phagocytosis and migration ability. Our results showed that Pb treatment significantly downregulated TREM2 expression and altered the localization of TREM2 expression in microglia. The protein expression of TREM2 was restored, and the inflammatory responses provoked by Pb exposure were ameliorated upon the overexpression of TREM2. Furthermore, the phagocytosis and migratory capabilities of microglia, which were impaired due to Pb exposure, were alleviated by TREM2 overexpression. Our in vitro findings were corroborated in vivo, demonstrating that TREM2 regulates the anti-inflammatory functions of microglia, thereby mitigating Pb-induced neuroinflammation. Our results provide insights into the detailed mechanism by which TREM2 alleviates Pb-induced neuroinflammation and suggest that activating the anti-inflammatory functions of TREM2 may represent a potential therapeutic strategy against environmental Pb-induced neurotoxicity.

Low-Complexity Multiple Transform Selection Combining Multi-Type Tree Partition Algorithm for Versatile Video Coding.

Despite the fact that Versatile Video Coding (VVC) achieves a superior coding performance to High-Efficiency Video Coding (HEVC), it takes a lot of time to encode video sequences due to the high computational complexity of the tools. Among these tools, Multiple Transform Selection (MTS) require the best of several transforms to be obtained using the Rate-Distortion Optimization (RDO) process, which increases the time spent video encoding, meaning that VVC is not suited to real-time sensor application networks. In this paper, a low-complexity multiple transform selection, combined with the multi-type tree partition algorithm, is proposed to address the above issue. First, to skip the MTS process, we introduce a method to estimate the Rate-Distortion (RD) cost of the last Coding Unit (CU) based on the relationship between the RD costs of transform candidates and the correlation between Sub-Coding Units' (sub-CUs') information entropy under binary splitting. When the sum of the RD costs of sub-CUs is greater than or equal to their parent CU, the RD checking of MTS will be skipped. Second, we make full use of the coding information of neighboring CUs to terminate MTS early. The experimental results show that, compared with the VVC, the proposed method achieves a 26.40% reduction in time, with a 0.13% increase in Bjøontegaard Delta Bitrate (BDBR).

Identification of key gene networks controlling vernalization development characteristics of Isatis indigotica by full-length transcriptomes and gene expression profiles.

Isatis indigotica Fort., as a common Chinese medicinal raw material, will lose its medicinal value if it blooms early, so it is highly valuable to clarify the induction mechanism of the vernalization of I. indigotica at low temperature. In this study, the concentrations of soluble sugar, proline, glutathione and zeatin in two germplasms of I. indigotica with different degrees of low temperature tolerance (Y1 and Y2) were determined at 10 days, 20 days and 30 days of low-temperature treatment, and the full-length transcriptome of 24 samples was sequenced by Nanopore sequencing with Oxford Nanopore Technologies (ONT). After that, the data of transcripts involved in the vernalization of I. indigotica at low temperature were obtained, and these transcripts were identified using weighted gene co-expression network analysis (WGCNA). The results revealed the massive accumulation of soluble sugar and proline in Y1 and Y2 after low temperature induction. A total of 18,385 new transcripts, 6168 transcription factors and 470 lncRNAs were obtained. Differential expression analysis showed that gibberellin, flavonoids, fatty acids and some processes related to low temperature response were significantly enriched. Eight key transcripts were identified by WGCNA, among which ONT.14640.1, ONT.9119.1, ONT.13080.2 and ONT.16007.1 encodes a flavonoid transporter, 9-cis-epoxycarotenoid dioxygenase 3 (NCED3), growth factor gene and L-aspartate oxidase in plants, respectively. It indicated that secondary metabolites such as hormones and flavonoids play an important role in the vernalization of I. indigotica. qRT-PCR proved the reliability of transcriptome results. These results provide important insights on the low-temperature vernalization of I. indigotica, and provide a research basis for analyzing the vernalization mechanism of I. indigotica. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12298-021-01110-2.

Linearized segmentation method for measuring the phase transfer function of a silicon Mach-Zehnder modulator by using the beat-frequency technique.

A linearity segmentation method for measuring the phase shift curve of silicon Mach-Zehnder modulators (SMZMs) as a function of applied electric field is presented. Applying a small sinusoidal signal to the traveling-wave electrode of the SMZM, the upper and lower arms of the SMZM produce differential phase modulation effect. Meanwhile, a local oscillator source with a wavelength-adjustable function is employed to heterodyne the intensity modulated optical signal of the SMZM; thus, the modulated intensity signal in the optical field domain is transformed into the low-frequency electric field domain. Meanwhile, a balanced detector with a low-speed transimpedance amplifier is exploited to realize photoelectric conversion, which can suppress the direct-current component and improve the anti-noise ability. Extracting the beat-frequency and first-order harmonic sideband signals in the case of phase bias is 0 and π, respectively, the phase shift slopes of the upper and lower arms can be calculated under different reverse PN voltages, and we can achieve the phase shift as a function of the modulation frequency and reverse PN voltage. The proposed method is supported by the simulation and measurement results and the key parameters of the SMZM, such as the radio frequency half-wave voltage, chirp characteristic, 3 dB bandwidth, etc., can be acquired from the phase shift curves of the upper and lower arms.

Thermo-optic-based phase-shifter power dither for silicon IQ optical modulator bias-control technology.

Recently, silicon optical in-phase quadrature (IQ) modulators have played an increasingly important role in coherent optical transmission networks because of their small package size and low cost. To stabilize the modulation performance of the silicon IQ optical modulator (SIQOM), the bias voltages of the SIQOM must be maintained at optimum points. Because of the nonlinear modulation characteristic of the silicon material, it is difficult to achieve high-precision closed-loop control of the bias voltage for the SIQOM. In this paper, a novel automatic bias-control scheme for the SIQOM is proposed and investigated theoretically and experimentally. First, two sinusoidal power dithers with different low frequencies are applied to the channels I and Q biases of the SIQOM. Next, a pair of orthogonal trigonometric functions with the same frequency as the power dither signal is constructed. We find that the optimum point of the bias voltage is the intersection of the orthogonal-integral curves via cross-correlation integral operations between the output signal of the SIQOM and the aforementioned trigonometric functions. The results indicate that the bias errors of the channels I/Q/P relative to the optimum point can be corrected precisely by the proposed scheme, and the jitters of the vector amplitude error caused by this scheme are <1% in 128-Gb/s dual-polarization quadrature phase-shift keying and single-polarization 16-quadrature amplitude modulation formats.

Study on auto bias control of a silicon optical modulator in a four-level pulse amplitude modulation format.

Recently, the high-speed silicon optical modulator has played a greater and greater role in optic fiber communication. Due to the silicon material nonlinear electro-optic effect, the quadrature (Quad) bias point of the modulator is difficult to be locked precisely. In this paper, a novel automatic bias control (ABC) method for four-level pulse amplitude modulation format silicon modulator is presented. First, a sinusoidal dither signal is applied to the modulator DC bias, and the normalized output optical signal function, which is modulated by the dither signal, is obtained. Then, a pair of orthogonal reference functions is created, whose numerical size is equal to the absolute value of sinusoidal and cosine functions so that the frequency is the same as the dither signal. Through the cross-correlation integral operation between the normalized and reference functions above, we find that the zero point of the above integral operation is the best Quad bias point through numerical simulation and experiment. The phase accuracy error of ABC is lower than 2 deg, and it satisfies the business specifications of a high-speed silicon optical modulator.

[Study on Genetic Stablity of Cryopreservation Regeneration Plant of Gentiana straminea with ISSR-PCR].

OBJECTIVE: To establish ISSR-PCR system of cryopreservation regeneration plant of Gentiana straminea, and to select appropriate primers and analyze the genetic stability. METHODS: DNA was extracted by CTAB, the optimal ISSR-PCR system was established by orthogonal experiment,and genetic stability was analyzed. RESULTS: The optimal ISSR-PCR system (25 µL) was established: dNTPs 0.50 µL, Mg2+ 1.00 µL, 10 x PCR Buffer 2.00 µL, primer 0.60 µL, Taq DNA polymerase 1.25 µL, template DNA 1.30 µL, and ddH2O 18.35 µL. The amplification program was devised: 94 degrees C for 5 min, denaturing at 94 degrees C for 30 s, annealing of 1 min due to denaturing temperature of different primer,extension at 72 degrees C for 1.5 min, 35 cycles, last extension at 72 degrees C for 7 min, conservation at 4 degrees C . The DNA mutation rate of cryopreservation regeneration plant of Gentiana straminea was 1.05%. CONCLUSION: The cryopreservation regeneration plant of Gentiana straminea retains very good genetic stability, there is little variation between each plant, so the cryopreservation can be used as a feasible method for resource protection of Gentiana straminea.

[Hyperbaric oxygenation and penis erectile dysfunction].

Relaxation and contraction factors influencing penile erection are produced and released by the central and peripheral nerves as well as intracavernosal sinus gap and vascular endothelial cells. Aging, diabetes, cardiovascular disease and spinal cord injury can influence these factors. Further researches of hyperbaric oxygenation (HBO) on the erectile dysfunction (ED) can provide some theoretical evidences for the clinical treatment of ED.

HCT: Chinese Medical Machine Reading Comprehension Question-Answering via Hierarchically Collaborative Transformer.

Chinese medical machine reading comprehension question-answering (cMed-MRCQA) is a critical component of the intelligence question-answering task, focusing on the Chinese medical domain question-answering task. Its purpose enable machines to analyze and understand the given text and question and then extract the accurate answer. To enhance cMed-MRCQA performance, it is essential to possess a profound comprehension and analysis of the context, deduce concealed information from the textual content and, subsequently, precisely determine the answer's span. The answer span has predominantly been defined by language items, with sentences employed in most instances. However, it is worth noting that sentences may not be properly split to varying degrees in various languages, making it challenging for the model to predict the answer zone. To alleviate this issue, this paper presents a novel architecture called HCT based on a Hierarchically Collaborative Transformer. Specifically, we presented a hierarchical collaborative method to locate the boundaries of sentence and answer spans separately. First, we designed a hierarchical encoding module to obtain the local semantic features of the corpus; second, we proposed a sentence-level self-attention module and a fused interaction-attention module to get the global information about the text. Finally, the model is trained by combining loss functions. Extensive experiments were conducted on the public dataset CMedMRC and the reconstruction dataset eMedicine to validate the effectiveness of the proposed method. Experimental results showed that the proposed method performed better than the state-of-the-art methods. Using the F1 metric, our model scored 90.4% on the CMedMRC and 73.2% on eMedicine.

Cholesterol-lowering effects of rhubarb free anthraquinones and their mechanism of action.

Rhubarb free anthraquinones (RhA) have significant lipid-regulating activity. However, whether RhA monomers have a role in lipid-regulating and their mechanism of action remains unclear. Based on the cholesterol accumulated HepG2 cell model, the cholesterol-regulating effect of RhA monomers and their combinations was investigated. The expression of sterol-regulatory element binding protein 2 (SREBP2), 3-hydroxy-3-methyl glutaryl coenzyme A reductase (HMGCR) and squalene monooxygenase (SQLE) of the model cells was analyzed to preliminarily explore the mechanism of action. After that, the liposomes of each active RhA monomer were separately prepared with the same lipid materials and the same preparation method so that each monomer has similar or equal bioavailability after oral administration to rats. Finally, the hypercholesterolemic rat model was established, and the effect of active RhA monomers loaded liposomes as well as their combinations on cholesterol-regulating was investigated and their mechanism of action was analyzed. The results showed that aloe-emodin, rhein and emodin were the main cholesterol-regulating components of RhA, and the combination of rhein and emodin showed significant cholesterol-lowering effect, which may be related to the expression of SREBP2, HMGCR and SQLE in the rat liver.

Manganese-induced miR-125b-2-3p promotes anxiety-like behavior via TFR1-mediated ferroptosis.

The toxic effects of excessive manganese (Mn) levels in the environment have led to a severe public health concern. Ferroptosis is a newly form of cell death relying on iron, inherent to pathophysiological processes of psychiatric disorders, such as anxiety and depression-like behaviors. Excessive Mn exposure causes various neurological effects, including neuronal death and mood disorders. Whether Mn exposure causes anxiety and depression-like behaviors, and the underlying mechanisms of Mn-induced ferroptosis have yet to be determined. Here, Mn-exposed mice showed anxiety-like behavior. We also confirmed the accumulation of ferrous ion (Fe2+), lipid peroxidation, and depletion of antioxidant defense system both in vitro and in vivo Mn-exposed models, suggesting that Mn exposure can induce ferroptosis. Furthermore, Mn exposure downregulated the expression of miR-125b-2-3p. In turn, overexpression of miR-125b-2-3p alleviated the Mn-induced ferroptosis by targeting Transferrin receptor protein 1 (TFR1). In summary, this novel study established the propensity of Mn to cause anxiety-like behavior, an effect that was regulated by miR-125b-2-3p and ensuing ferroptosis secondary to the targeting of TFR1. These results offer promising targets for the prevention and treatment of Mn-induced neurotoxicity.

[Study on rapid propagation of Scutellaria baicalensis variation].

OBJECTIVE: To establish and optimize the technology of taking root and promoting seedlings of white flower Scutellaria baicalensis test tube plantlet, and provide the theory and technology base for efficient factorization production system of white flower Scutellaria baicalensis. METHODS: Stem segments with axillary bud were cultured onto the different basic medium with different kinds and concentration of cytokinin and auxins to take root and produce strong seedling. RESULTS: The suitable culture medium for taking root of white flower Scutellaria baicalensis was 1/2 MS (all substance reduced half) + IBA 0.02 mg/L + sucrose 2%, the induction rate of root was 100%; The best medium for promoting seedling was 1/2 MS (all substance reduced half) + PP333 0.2 mg/L + IBA 0.02 mg/L + sucrose 2%, the seedling was green, the internode was normal, and its growth was vigorous and healthy. CONCLUSION: 1/2 MS (all substance reduced half) culture medium and relatively low concentration of sucrose is beneficial to inducing roots; Media adding appropriate concentration of IBA can significantly increase the root induction rate, the seedling has many stout roots; PP333 has dwarfing effect on seedlings and suitable concentration of PP333 can significantly improve the quality of the plantlets. A good technology of taking root and producing strong test tube plantlets is established.

[Comparison of content of total alkaloids and aconitine in Aconitum szechenyianum from different producing areas on basis of macroporous adsorption resin purification technology].

OBJECTIVE: To optimize the conditions of purifying the total alkaloids in Aconitum szechenyianum with macroporous adsorption resin, and compare the content of total alkaloids and aconitine in A. szechenyianum from different producing areas, in order to provide basis for further studies. METHOD: The orthogonal experiment method was adopted for optimizing the conditions of purifying the total alkaloids in A. szechenyianum with macroporous adsorption resin. The content of total alkaloids and aconitine were determined by using the titration method. The total alkaloids in A. szechenyianum from different producing areas were purified under optimum processing conditions. Aconitine was determined by HPLC. RESULT: Different processing conditions showed different influences on the purification of total alkaloids. The optimum conditions were resin type HPD-722, ethanol concentration of 80% , and ethanol elution volume of 80 mL x min(-1). The contents of aconitine in A. szechenyianum from different producing areas--Qinghai, Maxianshan, Ningxia and Yongdeng were 0.493 5, 0.883 5, 1.527 8, 1.664 4 mg x g(-1), respectively. CONCLUSION: The optimum processing conditions used in this essay could be used for purifying the total alkaloids and aconitine. A. szechenyianum from Yongdeng and Gansu contains the highest content of aconitine.

MAGE: Multi-scale Context-aware Interaction based on Multi-granularity Embedding for Chinese Medical Question Answer Matching.

BACKGROUND AND OBJECTIVE: The Chinese medical question answer matching (cMedQAM) task is the essential branch of the medical question answering system. Its goal is to accurately choose the correct response from a pool of candidate answers. The relatively effective methods are deep neural network-based and attention-based to obtain rich question-and-answer representations. However, those methods overlook the crucial characteristics of Chinese characters: glyphs and pinyin. Furthermore, they lose the local semantic information of the phrase by generating attention information using only relevant medical keywords. To address this challenge, we propose the multi-scale context-aware interaction approach based on multi-granularity embedding (MAGE) in this paper. METHODS: We adapted ChineseBERT, which integrates Chinese characters glyphs and pinyin information into the language model and fine-tunes the medical corpus. It solves the common phenomenon of homonyms in Chinese. Moreover, we proposed a context-aware interactive module to correctly align question and answer sequences and infer semantic relationships. Finally, we utilized the multi-view fusion method to combine local semantic features and attention representation. RESULTS: We conducted validation experiments on the three publicly available datasets, namely cMedQA V1.0, cMedQA V2.0, and cEpilepsyQA. The proposed multi-scale context-aware interaction approach based on the multi-granularity embedding method is validated by top-1 accuracy. On cMedQA V1.0, cMedQA V2.0, and cEpilepsyQA, the top-1 accuracy on the test dataset was improved by 74.1%, 82.7%, and 60.9%, respectively. Experimental results on the three datasets demonstrate that our MAGE achieves superior performance over state-of-the-art methods for the Chinese medical question answer matching tasks. CONCLUSIONS: The experiment results indicate that the proposed model can improve the accuracy of the Chinese medical question answer matching task. Therefore, it may be considered a potential intelligent assistant tool for the future Chinese medical answer question system.

Deep learning method of stochastic reconstruction of three-dimensional digital cores from a two-dimensional image.

Digital cores can characterize the true internal structure of rocks at the pore scale. This method has become one of the most effective ways to quantitatively analyze the pore structure and other properties of digital cores in rock physics and petroleum science. Deep learning can precisely extract features from training images for a rapid reconstruction of digital cores. Usually, the reconstruction of three-dimensional (3D) digital cores is performed by optimization using generative adversarial networks. The training data required for the 3D reconstruction are 3D training images. In practice, two-dimensional (2D) imaging devices are widely used because they can achieve faster imaging, higher resolution, and easier identification of different rock phases, so replacing 3D images with 2D ones avoids the difficulty of acquiring 3D images. In this paper, we propose a method, named EWGAN-GP, for the reconstruction of 3D structures from a 2D image. Our proposed method includes an encoder, a generator, and three discriminators. The main purpose of the encoder is to extract statistical features of a 2D image. The generator extends the extracted features into 3D data structures. Meanwhile, the three discriminators have been designed to gauge the similarity of morphological characteristics between cross sections of the reconstructed 3D structure and the real image. The porosity loss function is used to control the distribution of each phase in general. In the entire optimization process, a strategy using Wasserstein distance with gradient penalty makes the convergence of the training process faster and the reconstruction result more stable; it also avoids the problems of gradient disappearance and mode collapse. Finally, the reconstructed 3D structure and the target 3D structure are visualized to ascertain their similar morphologies. The morphological parameter indicators of the reconstructed 3D structure were consistent with those of the target 3D structure. The microstructure parameters of the 3D structure were also compared and analyzed. The proposed method can achieve accurate and stable 3D reconstruction compared with classical stochastic methods of image reconstruction.

Multi-modal cross-attention network for Alzheimer's disease diagnosis with multi-modality data.

Alzheimer's disease (AD) is a neurodegenerative disorder, the most common cause of dementia, so the accurate diagnosis of AD and its prodromal stage mild cognitive impairment (MCI) is significant. Recent studies have demonstrated that multiple neuroimaging and biological measures contain complementary information for diagnosis. Many existing multi-modal models based on deep learning simply concatenate each modality's features despite substantial differences in representation spaces. In this paper, we propose a novel multi-modal cross-attention AD diagnosis (MCAD) framework to learn the interaction between modalities for better playing their complementary roles for AD diagnosis with multi-modal data including structural magnetic resonance imaging (sMRI), fluorodeoxyglucose-positron emission tomography (FDG-PET) and cerebrospinal fluid (CSF) biomarkers. Specifically, the imaging and non-imaging representations are learned by the image encoder based on cascaded dilated convolutions and CSF encoder, respectively. Then, a multi-modal interaction module is introduced, which takes advantage of cross-modal attention to integrate imaging and non-imaging information and reinforce relationships between these modalities. Moreover, an extensive objective function is designed to reduce the discrepancy between modalities for effectively fusing the features of multi-modal data, which could further improve the diagnosis performance. We evaluate the effectiveness of our proposed method on the ADNI dataset, and the extensive experiments demonstrate that our MCAD achieves superior performance for multiple AD-related classification tasks, compared to several competing methods. Also, we investigate the importance of cross-attention and the contribution of each modality to the diagnostics performance. The experimental results demonstrate that combining multi-modality data via cross-attention is helpful for accurate AD diagnosis.