Search

Deep learning networks have yielded promising insights in the field of image classification. However, the hierarchical image classification (HIC) task, which involves assigning multiple, hierarchically organized labels to each image, presents a notable challenge. In response to this complexity, we developed a novel framework (HCAM-CL), which integrates a hierarchical cross-attention mechanism with a CNN-LSTM architecture for the HIC task. The HCAM-CL model effectively identifies the relevance between images and their corresponding labels while also being attuned to learning the hierarchical inter-dependencies among labels. Our versatile model is designed to manage both fixed-length and variable-length classification pathways within the hierarchy. In the HCAM-CL model, the CNN module is responsible for the essential task of extracting image features. The hierarchical cross-attention mechanism vertically aligns these features with hierarchical levels, uniformly weighing the importance of different spatial regions. Ultimately, the LSTM module is strategically utilized to generate predictive outcomes by treating HIC as a sequence generation challenge. Extensive experimental evaluations on CIFAR-10, CIFAR-100, and design patent image datasets demonstrate that our HCAM-CL framework consistently outperforms other state-of-the-art methods in hierarchical image classification. [ABSTRACT FROM AUTHOR]

Simple Summary: This study explored the prognostic significance of the pontine-white matter (PW) score in primary central nervous system (CNS) lymphoma patients with post-treatment 18F-FDG PET/CT and PET/MR imaging. Eligible patients were enrolled from January 2014 to December 2022. The PW score, derived from FDG uptake of the pons and white matter, was used to evaluate the metabolic activity of the treated lesion and its prognostic implications. A total of 90 patients across PET/CT and PET/MR modalities were assessed. The PW score demonstrated a robust discriminative ability in identifying patients with worse outcomes. It was also found to be a significant and independent indicator for worse prognosis in both PET/CT and PET/MR groups. The study demonstrated that this novel internal standardization indicator was an effective tool for risk stratification in primary CNS lymphoma post-treatment scenarios. Background: Limited data exist on the significance of PET imaging and quantitative PET parameters in primary central nervous system (CNS) lymphoma due to its relative rarity. This study was conducted to investigate the prognostic value of a novel internal standardization indicator, the pontine-white matter (PW) score, in primary CNS lymphoma patients undergoing post-treatment 18F-FDG PET/CT and PET/MR imaging. Methods: From January 2014 to December 2022, eligible patients with primary CNS lymphoma who underwent post-treatment PET imaging were enrolled. Using the FDG uptake of the pons and white matter as an internal reference, the PW score was graded based on the metabolism of the post-therapeutic lesion for each patient, and its associations with patients' prognosis were investigated. Results: In total, 41 patients with post-treatment PET/CT and 49 patients with post-treatment PET/MR imaging were enrolled. ROC curve analysis indicated that the PW score possessed robust discriminative ability in distinguishing patients with worse outcomes. Furthermore, a higher PW score was significantly correlated with and identified as an independent prognostic indicator for, worse prognosis in both the PET/CT and PET/MR cohorts. Conclusion: The study demonstrated that the PW score was an effective prognostic indicator for identifying post-treatment primary CNS lymphoma patients with worse outcomes. [ABSTRACT FROM AUTHOR]

Assume that (Y , ρ) is a nontrivial complete metric space, and that (Y , g 1 , ∞) is a time-varying discrete dynamical system (T-VDDS), which is given by sequences (g l) l = 1 ∞ of continuous selfmaps g l : Y → Y . In this paper, for a given Furstenberg family G and a given T-VDDS (Y , g 1 , ∞) , G -scrambled pairs of points of the system (Y , g 1 , ∞) (which contains the well-known scrambled pairs) are provided. Some properties of the set of G -scrambled pairs of a given T-VDDS (Y , g 1 , ∞) are studied. Moreover, the generically G -chaotic T-VDDS and the generically strongly G -chaotic T-VDDS are defined. A sufficient condition for a given T-VDDS to be generically strongly G -chaotic is also presented. [ABSTRACT FROM AUTHOR]

Chromosomal rearrangements often occur at genomic loci with DNA secondary structures, such as common fragile sites (CFSs) and palindromic repeats. We developed assays in mammalian cells that revealed CFS-derived AT-rich sequences and inverted Alu repeats (Alu-IRs) are mitotic recombination hotspots, requiring the repair functions of carboxy-terminal binding protein (CtBP)-interacting protein (CtIP) and the Mre11/Rad50/Nbs1 complex (MRN). We also identified an endonuclease activity of CtIP that is dispensable for end resection and homologous recombination (HR) at I-SceI-generated "clean" double-strand breaks (DSBs) but is required for repair of DSBs occurring at CFS-derived AT-rich sequences. In addition, CtIP nuclease-defective mutants are impaired in Alu-IRs-induced mitotic recombination. These studies suggest that an end resection-independent CtIP function is important for processing DSB ends with secondary structures to promote HR. Furthermore, our studies uncover an important role of MRN, CtIP, and their associated nuclease activities in protecting CFSs in mammalian cells.

Microhomology-mediated end joining (MMEJ) is a major pathway for Ku-independent alternative nonhomologous end joining, which contributes to chromosomal translocations and telomere fusions, but the underlying mechanism of MMEJ in mammalian cells is not well understood. In this study, we demonstrated that, distinct from Ku-dependent classical nonhomologous end joining, MMEJ--even with very limited end resection--requires cyclin-dependent kinase activities and increases significantly when cells enter S phase. We also showed that MMEJ shares the initial end resection step with homologous recombination (HR) by requiring meiotic recombination 11 homolog A (Mre11) nuclease activity, which is needed for subsequent recruitment of Bloom syndrome protein (BLM) and exonuclease 1 (Exo1) to DNA double-strand breaks (DSBs) to promote extended end resection and HR. MMEJ does not require S139-phosphorylated histone H2AX (γ-H2AX), suggesting that initial end resection likely occurs at DSB ends. Using a MMEJ and HR competition repair substrate, we demonstrated that MMEJ with short end resection is used in mammalian cells at the level of 10-20% of HR when both HR and nonhomologous end joining are available. Furthermore, MMEJ is used to repair DSBs generated at collapsed replication forks. These studies suggest that MMEJ not only is a backup repair pathway in mammalian cells, but also has important physiological roles in repairing DSBs to maintain cell viability, especially under genomic stress.