FLUX-Text: A Simple and Advanced Diffusion Transformer Baseline for Scene Text Editing

ArXiv.org · 2025-05-06

Scene text editing aims to modify or add texts on images while ensuring text fidelity and overall visual quality consistent with the background. Recent methods are primarily built on UNet-based diffusion models, which have improved scene text editing results, but still struggle with complex glyph structures, especially for non-Latin ones (\eg, Chinese, Korean, Japanese). To address these issues, we present \textbf{FLUX-Text}, a simple and advanced multilingual scene text editing DiT method. Specifically, our FLUX-Text enhances glyph understanding and generation through lightweight Visual and Text Embedding Modules, while preserving the original generative capability of FLUX. We further propose a Regional Text Perceptual Loss tailored for text regions, along with a matching two-stage training strategy to better balance text editing and overall image quality. Benefiting from the DiT-based architecture and lightweight feature injection modules, FLUX-Text can be trained with only $0.1$M training examples, a \textbf{97\%} reduction compared to $2.9$M required by popular methods. Extensive experiments on multiple public datasets, including English and Chinese benchmarks, demonstrate that our method surpasses other methods in visual quality and text fidelity. All the code is available at https://github.com/AMAP-ML/FluxText.

Alzheimer’s disease patient brain extracts induce multiple pathologies in novel vascularized neuroimmune organoids for disease modeling and drug discovery

Molecular Psychiatry · 2025-05-02 · 27 citations

Alzheimer's Disease (AD) is the most common cause of dementia, afflicting 55 million individuals worldwide, with limited treatment available. Current AD models mainly focus on familial AD (fAD), which is due to genetic mutations. However, models for studying sporadic AD (sAD), which represents over 95% of AD cases without specific genetic mutations, are severely limited. Moreover, the fundamental species differences between humans and animals might significantly contribute to clinical failures for AD therapeutics that have shown success in animal models, highlighting the urgency to develop more translational human models for studying AD, particularly sAD. In this study, we developed a complex human pluripotent stem cell (hPSC)-based vascularized neuroimmune organoid model, which contains multiple cell types affected in human AD brains, including human neurons, microglia, astrocytes, and blood vessels. Importantly, we demonstrated that brain extracts from individuals with sAD can effectively induce multiple AD pathologies in organoids four weeks post-exposure, including amyloid beta (Aβ) plaque-like aggregates, tau tangle-like aggregates, neuroinflammation, elevated microglial synaptic pruning, synapse/neuronal loss, and impaired neural network activity. Proteomics analysis also revealed disrupted AD-related pathways in our vascularized AD neuroimmune organoids. Furthermore, after treatment with Lecanemab, an FDA-approved antibody drug targeting Aβ, AD brain extracts exposed organoids showed a significant reduction of amyloid burden, along with an elevated vascular inflammation response. Thus, the vascularized neuroimmune organoid model provides a unique opportunity to study AD, particularly sAD, under a pathophysiological relevant three-dimensional (3D) human cell environment. It also holds great promise to facilitate AD drug development, particularly for immunotherapies.

Human IPSC-Derived Microglia Sense and Dampen Hyperexcitability of Cortical Neurons Carrying the Epilepsy-Associated<i>SCN2A</i>-L1342P Mutation

Journal of Neuroscience · 2024-11-18 · 11 citations

Neuronal hyperexcitability is a hallmark of epilepsy. It has been recently shown in rodent models of seizures that microglia, the brain's resident immune cells, can respond to and modulate neuronal excitability. However, how human microglia interact with human neurons to regulate hyperexcitability mediated by an epilepsy-causing genetic mutation found in patients is unknown. The SCN2A gene is responsible for encoding the voltage-gated sodium channel Nav1.2, one of the leading contributors to monogenic epilepsies. Previously, we demonstrated that the recurring Nav1.2-L1342P mutation leads to hyperexcitability in a male donor (KOLF2.1) human-induced pluripotent stem cell (hiPSC)-derived cortical neuron model. Microglia originate from a different lineage (yolk sac) and are not naturally present in hiPSC-derived neuronal cultures. To study how microglia respond to neurons carrying a disease-causing mutation and influence neuronal excitability, we established a coculture model comprising hiPSC-derived neurons and microglia. We found that microglia display increased branch length and enhanced process-specific calcium signal when cocultured with Nav1.2-L1342P neurons. Moreover, the presence of microglia significantly lowered the repetitive action potential firing and current density of sodium channels in neurons carrying the mutation. Additionally, we showed that coculturing with microglia led to a reduction in sodium channel expression within the axon initial segment of Nav1.2-L1342P neurons. Furthermore, we demonstrated that Nav1.2-L1342P neurons release a higher amount of glutamate compared with control neurons. Our work thus reveals a critical role of human iPSC-derived microglia in sensing and dampening hyperexcitability mediated by an epilepsy-causing mutation.

Emerging Human Pluripotent Stem Cell-Based Human–Animal Brain Chimeras for Advancing Disease Modeling and Cell Therapy for Neurological Disorders

Neuroscience Bulletin · 2024-03-11 · 6 citations

Microglial over-pruning of synapses during development in autism-associated SCN2A-deficient mice and human cerebral organoids

Molecular Psychiatry · 2024-03-18 · 54 citations

Alzheimer’s Disease Patient Brain Extracts Induce Multiple Pathologies in Vascularized Neuroimmune Organoids for Disease Modeling and Drug Discovery

bioRxiv (Cold Spring Harbor Laboratory) · 2024-10-28 · 3 citations

Abstract Alzheimer’s Disease (AD) is the most common cause of dementia afflicting 55 million individuals worldwide, with limited treatment available. Current AD models mainly focus on familial AD (fAD), which is due to genetic mutations. However, models for studying sporadic AD (sAD), which represents over 95% of AD cases without specific genetic mutations, are severely limited. Moreover, the fundamental species differences between humans and animals might significantly contribute to clinical failures for AD therapeutics that have shown success in animal models, highlighting the urgency to develop more translational human models for studying AD, particularly sAD. In this study, we developed a complex human pluripotent stem cell (hPSC)-based vascularized neuroimmune organoid model, which contains multiple cell types affected in human AD brains, including human neurons, microglia, astrocytes, and blood vessels. Importantly, we demonstrated that brain extracts from individuals with sAD can effectively induce multiple AD pathologies in organoids four weeks post-exposure, including amyloid beta (Aβ) plaques-like aggregates, tau tangles-like aggregates, neuroinflammation, elevated microglial synaptic pruning, synapse/neuronal loss, and impaired neural network. Furthermore, after treatment with Lecanemab, an FDA-approved drug targeting Aβ, AD brain extract exposed organoids showed a significant reduction of amyloid burden. Thus, the neuroimmune organoid model provides a unique opportunity to study AD, particularly sAD under a pathophysiological relevant three-dimensional (3D) human cell environment. It also holds great promise to facilitate AD drug development, particularly for immunotherapies.

Develop a vascularized neuroimmune organoid model for studying sporadic Alzheimer’s disease

Alzheimer s & Dementia · 2024-12-01

BACKGROUND: Alzheimer's disease (AD) is the leading cause of dementia, affecting 50 million people globally. Current AD animal models mainly focus on familial or inherited AD. These models often carry the APP and PSEN gene mutations from familial AD patients, or introduce microtubule-associated protein tau (MAPT) mutations, which can cause frontotemporal dementia but are not linked to AD. Notably, there are significant species differences between humans and animals, and most AD therapeutics that have shown success in animal models have failed in clinical, highlighting the development of more translational human-centric models for studying AD, particularly sporadic AD. METHOD: In the current study, we developed a human-induced pluripotent stem cell-based vascularized neuroimmune organoids that contain multiple cell types in the human brain, including neurons, microglia, astrocytes, and blood vessels. We further challenge brain organoids with human AD postmortem brain tissues-derived extracts to induce AD pathologies for studying sporadic AD. Finally, we applied this new model to test the effect of AD immunotherapy, such as Lecanemab. RESULT: AD brain extracts that contain pathological seeds, such as amyloid and tau, successfully induce multiple AD pathologies in neuro-immune organoids, including 6E10+Aβ plagues, AT8+tau tangles, as well as neuroinflammation. CONCLUSION: The neuroimmune organoid model provides a unique opportunity to model sporadic AD in a human cell setting. This innovative model also facilitates AD drug screening, particularly for testing AD immunotherapy, such as antibody-based AD treatment.

Microglial over-pruning of synapses during development in autism-associated SCN2A-deficient mice and human cerebral organoids

Research Square · 2023-09-28 · 15 citations

CCDC 2234062: Experimental Crystal Structure Determination

The Cambridge Structural Database · 2023-01-19

An entry from the Cambridge Structural Database, the world’s repository for small molecule crystal structures. The entry contains experimental data from a crystal diffraction study. The deposited dataset for this entry is freely available from the CCDC and typically includes 3D coordinates, cell parameters, space group, experimental conditions and quality measures.

CCDC 2234060: Experimental Crystal Structure Determination

The Cambridge Structural Database · 2023-01-19

An entry from the Cambridge Structural Database, the world’s repository for small molecule crystal structures. The entry contains experimental data from a crystal diffraction study. The deposited dataset for this entry is freely available from the CCDC and typically includes 3D coordinates, cell parameters, space group, experimental conditions and quality measures.