Semi-quantitative scoring criteria based on multiple staining methods combined with machine learning to evaluate residual nuclei in decellularized matrix DOI Creative Commons

Meng Zhong,

Hongwei He,

Panxianzhi Ni

et al.

Regenerative Biomaterials, Journal Year: 2024, Volume and Issue: 12

Published: Dec. 18, 2024

Abstract The detection of residual nuclei in decellularized extracellular matrix (dECM) biomaterials is critical for ensuring their quality and biocompatibility. However, current evaluation methods have limitations addressing impurity interference providing intelligent analysis. In this study, we utilized four staining techniques—hematoxylin-eosin staining, acetocarmine the Feulgen reaction 4’,6-diamidino-2-phenylindole staining—to detect dECM biomaterials. Each method was quantitatively evaluated across multiple parameters, including area, perimeter grayscale values, to establish a semi-quantitative scoring system nuclei. These quantitative data were further employed as learning indicators machine models designed automatically identify experimental results demonstrated that no single alone could accurately differentiate between impurities. table developed. With table, accuracy determining whether suspicious point cell nucleus has reached over 98%. By combining methods, false positives caused by contamination eliminated. automatic recognition model trained based on nuclear parameter features optimal index after several iterations training 172 epochs. artificial intelligence achieved 90% detecting use multidimensional integrated with learning, significantly improved identifying residues slices. This approach provides more reliable objective evaluating biomaterials, while also increasing efficiency.

Language: Английский

Benchmarking pathology foundation models for non-neoplastic pathology in the placenta DOI Creative Commons

Zhongyuan Peng,

Marina A. Ayad,

Jing You

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: March 20, 2025

Machine learning (ML) applications within diagnostic histopathology have been extremely successful. While many successful models built using general-purpose trained largely on everyday objects, there is a recent trend toward pathology-specific foundation models, images. Pathology show strong performance cancer detection and subtyping, grading, predicting molecular diagnoses. However, we noticed lacunae in the testing of models. Nearly all benchmarks used to test them are focused cancer. Neoplasia an important pathologic mechanism key concern much clinical pathology, but it represents one bases disease. Non-neoplastic pathology dominates findings placenta, critical organ human development, as well specimen commonly encountered practice. Very little none data training placenta. Thus, placental doubly out distribution, representing useful challenge for We developed estimation gestational age, classifying normal tissue, identifying inflammation umbilical cord membranes, classification macroscopic lesions including villous infarction, intervillous thrombus, perivillous fibrin deposition. tested 5 4 non-pathology each benchmark tasks zero-shot K-nearest neighbor regression, content-based image retrieval, supervised whole-slide attention-based multiple instance learning. In task, best performing model was model. gap between diminished related or those which task performed embeddings. Performance comparable among Among ResNet consistently worse, while from present decade showed better performance. Future work could examine impact incorporating into training.

Language: Английский

Citations

0
Artificial Intelligence in Placental Pathology: New Diagnostic Imaging Tools in Evolution and in Perspective DOI Creative Commons
Antonio d’Amati, Giorgio Maria Baldini,

Tommaso Difonzo

et al.

Journal of Imaging, Journal Year: 2025, Volume and Issue: 11(4), P. 110 - 110

Published: April 3, 2025

Artificial intelligence (AI) has emerged as a transformative tool in placental pathology, offering novel diagnostic methods that promise to improve accuracy, reduce inter-observer variability, and positively impact pregnancy outcomes. The primary objective of this review is summarize recent developments AI applications tailored specifically histopathology. Current AI-driven approaches include advanced digital image analysis, three-dimensional reconstruction, deep learning models such GestAltNet for precise gestational age estimation automated identification histological lesions, including decidual vasculopathy maternal vascular malperfusion. Despite these advancements, significant challenges remain, notably dataset heterogeneity, interpretative limitations current algorithms, issues regarding model transparency. We critically address by proposing targeted solutions, augmenting training datasets with annotated artifacts, promoting explainable methods, enhancing cross-institutional collaborations. Finally, we outline future research directions, emphasizing the refinement algorithms routine clinical integration fostering interdisciplinary cooperation among pathologists, computational researchers, specialists.

Language: Английский

Citations

0
Deep Learning for Fetal Inflammatory Response Diagnosis in the Umbilical Cord DOI

Marina A. Ayad,

Ramin Nateghi,

Abhishek Sharma

et al.

Placenta, Journal Year: 2025, Volume and Issue: 167, P. 1 - 10

Published: April 24, 2025

Language: Английский

Citations

0
Trends and Challenges of the Modern Pathology Laboratory for Biopharmaceutical Research Excellence DOI Creative Commons
Sílvia Sisó, Anoop Kavirayani, Suzana S. Couto

et al.

Toxicologic Pathology, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 13, 2024

Pathology, a fundamental discipline that bridges basic scientific discovery to the clinic, is integral successful drug development. Intrinsically multimodal and multidimensional, anatomic pathology continues be empowered by advancements in molecular digital technologies enabling spatial tissue detection of biomolecules such as genes, transcripts, proteins. Over past two decades, breakthroughs biology automation digitization laboratory processes have enabled implementation higher throughput assays generation extensive data sets from sections biopharmaceutical research development units. It our goal provide readers with some rationale, advice, ideas help establish modern meet emerging needs research. This manuscript provides (1) high-level overview current state future vision for excellence practice (2) shared perspectives on how optimally leverage expertise discovery, toxicologic, translational pathologists effective spatial, molecular, support discovery. captures insights experiences, challenges, solutions laboratories various organizations, including their approaches troubleshooting adopting new technologies.

Language: Английский

Citations

0
Semi-quantitative scoring criteria based on multiple staining methods combined with machine learning to evaluate residual nuclei in decellularized matrix DOI Creative Commons

Meng Zhong,

Hongwei He,

Panxianzhi Ni

et al.

Regenerative Biomaterials, Journal Year: 2024, Volume and Issue: 12

Published: Dec. 18, 2024

Abstract The detection of residual nuclei in decellularized extracellular matrix (dECM) biomaterials is critical for ensuring their quality and biocompatibility. However, current evaluation methods have limitations addressing impurity interference providing intelligent analysis. In this study, we utilized four staining techniques—hematoxylin-eosin staining, acetocarmine the Feulgen reaction 4’,6-diamidino-2-phenylindole staining—to detect dECM biomaterials. Each method was quantitatively evaluated across multiple parameters, including area, perimeter grayscale values, to establish a semi-quantitative scoring system nuclei. These quantitative data were further employed as learning indicators machine models designed automatically identify experimental results demonstrated that no single alone could accurately differentiate between impurities. table developed. With table, accuracy determining whether suspicious point cell nucleus has reached over 98%. By combining methods, false positives caused by contamination eliminated. automatic recognition model trained based on nuclear parameter features optimal index after several iterations training 172 epochs. artificial intelligence achieved 90% detecting use multidimensional integrated with learning, significantly improved identifying residues slices. This approach provides more reliable objective evaluating biomaterials, while also increasing efficiency.

Language: Английский

Citations

0