Related Experiment Video
Updated: Jul 13, 2025
16:59
Computer-assisted Large-scale Visualization and Quantification of Pancreatic Islet Mass, Size Distribution and Architecture
Published on: March 4, 2011
12.3K
Jointly Optimized Spatial Histogram UNET Architecture (JOSHUA) for Adipose Tissue Segmentation.
Joshua K Peeples1, Julie F Jameson2, Nisha M Kotta3
1Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL 32611, USA.
BME Frontiers
|October 18, 2023
Summary
A new machine learning algorithm, JOSHUA, accurately quantifies adipose tissue in histological images. This method improves biomaterial assessment for soft tissue repair, offering a less biased alternative to manual analysis.
Area of Science:
- Biomaterials Science
- Computational Biology
- Histology
Background:
- Adipose tissue deposition is critical in biomaterial design for soft tissue repair.
- Current quantification methods are time-consuming and prone to bias.
- Silk fibroin biomaterials are being investigated for tissue regeneration.
Purpose of the Study:
- To develop a machine learning algorithm for quantifying adipose tissue at surgical sites.
- To apply convolutional neural network (CNN) models to histological images of silk fibroin implants.
- To establish a quantitative, unbiased method for evaluating biomaterial performance.
Main Methods:
- Utilized CNN models with novel spatial histogram layers for adipose tissue segmentation.
- Developed the Jointly Optimized Spatial Histogram UNET Architecture (JOSHUA) and its variant JOSHUA+.
- Compared JOSHUA models against baseline UNET and attention UNET models using H&E and Masson's trichrome stained images.
Main Results:
- JOSHUA models demonstrated improved performance in identifying and segmenting adipose tissue.
- Qualitative and quantitative evaluations confirmed the efficacy of histogram layers.
- The developed models accurately localized adipose tissue in histological samples.
Conclusions:
- JOSHUA and JOSHUA+ are effective tools for adipose tissue identification and localization in biomaterial studies.
- The proposed methods offer a significant advancement over traditional quantification techniques.
- A new histological dataset and code are publicly available to facilitate further research.
