Dartmouth Institute for Health Policy and Clinical Practice
Dartmouth Digital Commons (Dartmouth College)Not a member yet
8214 research outputs found
Sort by
Design and Validation of a Novel Patterned Microfluidic Culture Platform for hiPSC-CM Maturation
Full text linkAdult cardiomyocytes (CMs) are among the most challenging cells to culture in vitro, given their poor adherence to plastic culture wells and poor maturation outside the native extracellular matrix (Narkar et al., 2022). Human induced pluripotent stem cell-derived CMs (hiPSC-CMs) offer a viable alternative to study heart development and disease progression. While not comparable to adult CMs in their maturity (Narkar et al., 2022), they mimic the clinical phenotypes of heart failure and are more feasible to culture. Nevertheless, hiPSC-CMs require specialized culture conditions that cannot be achieved with 2D static culturing, which fail to replicate native tissue environments (X. Huang et al., 2022). Instead, microfluidic culturing may enhance the maturation profiles of hiPSC-CMs, given the ability to tune media flow rates to match cell shear stress values. Additionally, microfluidic culture may be enhanced through substrate micropatterning, which directs specific cell attachment toward the desired mature cell profiles. While previous research has outlined the impact of microfluidic media flow and micropatterning on hiPSC-CM maturation, little work has analyzed how dynamic media flow influences cells inside micropatterns, or how biomimetic-driven micropatterns can support maturation in combination with dynamic media flow. This work hypothesizes that dynamic flow combined with biomimetic-inspired micropatterns will improve hiPSC-CM maturation. Functional validation will include an analysis of cell morphology and gene expression compared to static culture. Through the intersection of advanced microfabrication and biological testing, this work will provide key insight into the impact of biomimetic culture on hiPSC-CM development, potentially establishing a new paradigm for cell culture
Unsafe2Safe: Controllable Image Anonymization for Downstream Utility
No full textLarge-scale image datasets frequently contain identifiable or sensitive content, raising privacy risks when training models that may memorize and leak such information. We present Unsafe2Safe, a fully automated pipeline that detects privacy-prone images and rewrites only their sensitive regions using multimodally guided diffusion editing. Unsafe2Safe operates in two stages. Stage 1 uses a vision--language model to (i) inspect images for privacy risks, (ii) generate paired private and public captions that respectively include and omit sensitive attributes, and (iii) prompt a large language model to produce structured, identity-neutral edit instructions conditioned on the public caption. Stage 2 employs instruction-driven diffusion editors to apply these dual textual prompts, producing privacy-safe images that preserve global structure and task-relevant semantics while neutralizing private content. To measure anonymization quality, we introduce a unified evaluation suite covering Quality, Cheating, and Utility dimensions. Across Caltech101 and MIT Indoor67, Unsafe2Safe reduces face similarity, text similarity, and demographic predictability by large margins, while maintaining downstream model accuracy comparable to training on raw data. Fine-tuning diffusion editors on our automatically generated triplets (private caption, public caption, edit instruction) further improves both privacy protection and semantic fidelity. Unsafe2Safe provides a scalable, principled solution for constructing large, privacy-safe datasets without sacrificing visual consistency or downstream utility
