A Cross-Domain Weakly Supervised Diabetic Retinopathy Lesion Identification Method Based on Multiple Instance Learning and Domain Adaptation.

Accurate identification of lesions and their use across different medical institutions are the foundation and key to the clinical application of automatic diabetic retinopathy (DR) detection. Existing detection or segmentation methods can achieve acceptable results in DR lesion identification, but they strongly rely on a large number of fine-grained annotations that are not easily accessible and suffer severe performance degradation in the cross-domain application. In this paper, we propose a cross-domain weakly supervised DR lesion identification method using only easily accessible coarse-grained lesion attribute labels. We first propose the novel lesion-patch multiple instance learning method (LpMIL), which leverages the lesion attribute label for patch-level supervision to complete weakly supervised lesion identification. Then, we design a semantic constraint adaptation method (LpSCA) that improves the lesion identification performance of our model in different domains with semantic constraint loss. Finally, we perform secondary annotation on the open-source dataset EyePACS, to obtain the largest fine-grained annotated dataset EyePACS-pixel, and validate the performance of our model on it. Extensive experimental results on the public dataset FGADR and our EyePACS-pixel demonstrate that compared with the existing detection and segmentation methods, the proposed method can identify lesions accurately and comprehensively, and obtain competitive results using only coarse-grained annotations.

Automatic Medical Report Generation Based on Cross-View Attention and Visual-Semantic Long Short Term Memorys.

Automatic medical report generation based on deep learning can improve the efficiency of diagnosis and reduce costs. Although several automatic report generation algorithms have been proposed, there are still two main challenges in generating more detailed and accurate diagnostic reports: using multi-view images reasonably and integrating visual and semantic features of key lesions effectively. To overcome these challenges, we propose a novel automatic report generation approach. We first propose the Cross-View Attention Module to process and strengthen the multi-perspective features of medical images, using mean square error loss to unify the learning effect of fusing single-view and multi-view images. Then, we design the module Medical Visual-Semantic Long Short Term Memorys to integrate and record the visual and semantic temporal information of each diagnostic sentence, which enhances the multi-modal features to generate more accurate diagnostic sentences. Applied to the open-source Indiana University X-ray dataset, our model achieved an average improvement of 0.8% over the state-of-the-art (SOTA) model on six evaluation metrics. This demonstrates that our model is capable of generating more detailed and accurate diagnostic reports.

Exergames improves cognitive functions in adolescents with depression: study protocol of a prospective, assessor-blind, randomized controlled trial.

BACKGROUND: Depression is a condition that imposes a significant disease burden, with cognitive impairment being one of its costly symptoms. While cognitive rehabilitation is crucial, it is also challenging. Although some studies have investigated the impact of exergames on cognitive function improvement, these have primarily focused on the elderly population, with limited attention given to individuals with depression. Consequently, this study aims to investigate the effects of exergames on cognitive functions in adolescents with depression and compare the effectiveness of exergames with traditional exercise. METHOD: The present investigation is a single-center randomized controlled trial that employs the ANOVA method to calculate the sample size using G*Power software, assuming a 25% dropout rate. The study enrolls fifty-four eligible patients with depression who are randomly allocated to one of three treatment groups: the exergames group, which receives standard treatment and exergames intervention; the exercise group, which receives standard treatment and traditional exercise intervention; and the control group, which receives standard treatment exclusively. The study provides a comprehensive regimen of 22 supervised exercise and exergame sessions over an 8-week period, with a frequency of twice per week for the initial two weeks and three times per week for the subsequent six weeks. The researchers gather cognitive, mood, and sleep metrics at the onset of the first week, as well as at the conclusion of the fourth and eighth weeks. The researchers employ a wearable device to track participants' heart rate during each intervention session and evaluate the Borg Rating of Perceived Exertion scale at the conclusion of each session. DISCUSSION: The findings from this study make several contributions to the current literature. First, this study comprehensively reports the efficacy of an exergames intervention for multidimensional symptoms in adolescents with depression. Second, this study also compares the efficacy of exergames with that of traditional exercise. These findings provide a theoretical basis for the use of exergames as an adjunctive intervention for depression and lay the groundwork for future research. TRIAL REGISTRATION: This trial is registered with the Chinese Clinical Trials Registry (Registration number: ChiCTR2100052709; Registration Status: Prospective registration;) 3/11/2021, URL:    http://www.chictr.org.cn/edit.aspx?pid=135663&htm=4 .

Mutations in viral nucleocapsid protein and endoRNase are discovered to associate with COVID19 hospitalization risk.

SARS-CoV-2 is spreading worldwide with continuously evolving variants, some of which occur in the Spike protein and appear to increase viral transmissibility. However, variants that cause severe COVID-19 or lead to other breakthroughs have not been well characterized. To discover such viral variants, we assembled a cohort of 683 COVID-19 patients; 388 inpatients ("cases") and 295 outpatients ("controls") from April to August 2020 using electronically captured COVID test request forms and sequenced their viral genomes. To improve the analytical power, we accessed 7137 viral sequences in Washington State to filter out viral single nucleotide variants (SNVs) that did not have significant expansions over the collection period. Applying this filter led to the identification of 53 SNVs that were statistically significant, of which 13 SNVs each had 3 or more variant copies in the discovery cohort. Correlating these selected SNVs with case/control status, eight SNVs were found to significantly associate with inpatient status (q-values < 0.01). Using temporal synchrony, we identified a four SNV-haplotype (t19839-g28881-g28882-g28883) that was significantly associated with case/control status (Fisher's exact p = 2.84 × 10-11). This haplotype appeared in April 2020, peaked in June, and persisted into January 2021. The association was replicated (OR = 5.46, p-value = 4.71 × 10-12) in an independent cohort of 964 COVID-19 patients (June 1, 2020 to March 31, 2021). The haplotype included a synonymous change N73N in endoRNase, and three non-synonymous changes coding residues R203K, R203S and G204R in the nucleocapsid protein. This discovery points to the potential functional role of the nucleocapsid protein in triggering "cytokine storms" and severe COVID-19 that led to hospitalization. The study further emphasizes a need for tracking and analyzing viral sequences in correlations with clinical status.

Nek8445, a protein kinase required for microtubule regulation and cytokinesis in Giardia lamblia.

Giardia has 198 Nek kinases whereas humans have only 11. Giardia has a complex microtubule cytoskeleton that includes eight flagella and several unique microtubule arrays that are utilized for parasite attachment and facilitation of rapid mitosis and cytokinesis. The need to regulate these structures may explain the parallel expansion of the number of Nek family kinases. Here we use live and fixed cell imaging to uncover the role of Nek8445 in regulating Giardia cell division. We demonstrate that Nek8445 localization is cell cycle regulated and this kinase has a role in regulating overall microtubule organization. Nek8445 depletion results in short flagella, aberrant ventral disk organization, loss of the funis, defective axoneme exit, and altered cell shape. The axoneme exit defect is specific to the caudal axonemes, which exit from the posterior of the cell, and this defect correlates with rounding of the cell posterior and loss of the funis. Our findings implicate a role for the funis in establishing Giardia's cell shape and guiding axoneme docking. On a broader scale our results support the emerging view that Nek family kinases have a general role in regulating microtubule organization.

Proxies of CRISPR/Cas9 Activity To Aid in the Identification of Mutagenized Arabidopsis Plants.

CRISPR/Cas9 has become the preferred gene-editing technology to obtain loss-of-function mutants in plants, and hence a valuable tool to study gene function. This is mainly due to the easy reprogramming of Cas9 specificity using customizable small non-coding RNAs, and to the possibility of editing several independent genes simultaneously. Despite these advances, the identification of CRISPR-edited plants remains time and resource-intensive. Here, based on the premise that one editing event in one locus is a good predictor of editing event/s in other locus/loci, we developed a CRISPR co-editing selection strategy that greatly facilitates the identification of CRISPR-mutagenized Arabidopsis thaliana plants. This strategy is based on targeting the gene/s of interest simultaneously with a proxy of CRISPR-Cas9-directed mutagenesis. The proxy is an endogenous gene whose loss-of-function produces an easy-to-detect visible phenotype that is unrelated to the expected phenotype of the gene/s under study. We tested this strategy via assessing the frequency of co-editing of three functionally unrelated proxy genes. We found that each proxy predicted the occurrence of mutations in each surrogate gene with efficiencies ranging from 68 to 100%. The selection strategy laid out here provides a framework to facilitate the identification of multiplex edited plants, thus aiding in the study of gene function when functional redundancy hinders the effort to define gene-function-phenotype links.

Myosin-independent cytokinesis in Giardia utilizes flagella to coordinate force generation and direct membrane trafficking.

Devoid of all known canonical actin-binding proteins, the prevalent parasite Giardia lamblia uses an alternative mechanism for cytokinesis. Unique aspects of this mechanism can potentially be leveraged for therapeutic development. Here, live-cell imaging methods were developed for Giardia to establish division kinetics and the core division machinery. Surprisingly, Giardia cytokinesis occurred with a median time that is ∼60 times faster than mammalian cells. In contrast to cells that use a contractile ring, actin was not concentrated in the furrow and was not directly required for furrow progression. Live-cell imaging and morpholino depletion of axonemal Paralyzed Flagella 16 indicated that flagella-based forces initiated daughter cell separation and provided a source for membrane tension. Inhibition of membrane partitioning blocked furrow progression, indicating a requirement for membrane trafficking to support furrow advancement. Rab11 was found to load onto the intracytoplasmic axonemes late in mitosis and to accumulate near the ends of nascent axonemes. These developing axonemes were positioned to coordinate trafficking into the furrow and mark the center of the cell in lieu of a midbody/phragmoplast. We show that flagella motility, Rab11, and actin coordination are necessary for proper abscission. Organisms representing three of the five eukaryotic supergroups lack myosin II of the actomyosin contractile ring. These results support an emerging view that flagella play a central role in cell division among protists that lack myosin II and additionally implicate the broad use of membrane tension as a mechanism to drive abscission.