Automated Detection and Diameter Estimation for Mouse Mesenteric Artery using Semantic Segmentation

Akinori Higaki, Ahmad U. M. Mahmoud, Pierre Paradis, Ernesto L. Schiffrin

Preprint posted on October 29, 2020

Improving the toolkit for vascular biology.

Selected by Mariana De Niz


Dysfunction of resistance vessels contributes to blood pressure elevation, and is an independent risk factor for cardiovascular events. Various studies have focused on the study of small mesenteric artery structure and function in hypertension and cardiovascular disease. Various methods, including pressurized myography, exist for the assessment of small arteries. However some of the procedures require high levels of expertise for sample preparation, and choice of artery size. Automatic vessel detection, with a size measurement system, could be helpful for vessel selection. In their work, Higaki et al (1) propose a novel workflow for image segmentation of mouse mesenteric arteries using U-net (a convolutional neural network developed for biomedical image segmentation).

Key findings and developments

Key methods

The study consisted on  collecting 654 mesenteric artery images from 59 mice from various genetic backgrounds (including fluorescent reporters and mutants). The mesenteric artery vascular bed attached to the intestine, was dissected from the abdominal cavity and collected in ice-cold PBS or Krebs solution. The upper surface of each mesenteric artery was exposed by dissecting out perivascular adipose tissue. Images were then collected using a surgical microscope. A principal component analysis (PCA) was performed for the characterization of the collected image. Segmented mask images were made using Labelme (a Python-based annotation tool), whereby each pixel was divided into 4 possible classes: background, arterial lumen, arterial wall, and off-target vessel (eg. veins or capillaries on the surface of the intestine). Arteries and veins were differentiated based on morphology. Additional to the segment information, vessel diameters were manually measured using ImageJ. U-Net architecture was used as a semantic segmentation model, with some modifications to fit specific needs. Segmented images were then used as ground truth to train the machine learning model. From the segmented output images, the vessel area and lumen area were obtained as binary images. Using the OpenCV python library, a total circumference length of the vessel contour and lumen contour were obtained. Based on a rectangular approximation, the circumference was calculated. The average vessel and lumen diameter were then obtained. The authors made this automated measurement system, publicly available at The agreement between the manual and automated methods was evaluated, as was the model performance for correct segmentation.

Descriptive analysis of the collected data

The majority of arteries were captured in the vertical orientation, while others were horizontally, or multi-oriented. Also, the majority of arteries were captured with the accompanying vein. A minority had bifurcations or side branches, and also a minority contained more than two arteries. Also a majority were surrounded by perivascular adipose tissue, while a small minority were accidentally damaged. In the collected data, the authors expand on the model performance for segmentation, and validated an automated method for line-width calculation by comparing it to the manual measurements. For both measurements, the automated calculation gave similar values to the manual measurements.

For vessel size and lumen size, there was a strong significant linear correlation between the true dimension and the estimated size based on the predicted contour. Finally, vessel contour before dissection and after pressurization at 45 mm Hg were assessed. The authors found no strong linear correlation of outer vessel or lumen size before and after pressurization. Width/length ratio before dissection had no linear correlation with pressurization. However, there was a strong correlation between width and length ratio before dissection and fold-change of lumen diameter.

The authors then aimed to construct multivariate binary logistic regression models to choose arteries, in terms of lumen size at the pressurized state. For the prediction upper limit, the vessel size and lumen size were chosen as explanatory variables, while for the prediction lower limit, it was the vessel size and the width/length ratio before dissection. This model showed 92.5% accuracy.

What I like about this preprint

I think novel methods for use in vascular biology are extremely useful to multiple fields of research (i.e. infectious diseases, cancer, cardiovascular diseases, neuronal diseases, etc). I think the authors generated and optimized an interesting method with interesting potential to various applications. Moreover, the authors made various of the methods they developed, fully available in open access.

Open questions

  1. Is the method hereby developed, equally applicable to other types of vasculature? For instance in the heart or the brain?
  2. You investigate two important parameters: isolation and pressurization. To what extent do you alter normal vascular pressurization upon isolation from the surrounding tissue (eg. adipose tissue)?
  3. You describe at the beginning of your work, the 4 different classifications of pixels, as well as the percentage of vessels showing specific characteristics, including branching. How does branching affect the automated segmentation?
  4. Following from question 1, are the training data generated from a specific vessel type applicable to other vessels of interest?


  1. Higaki et al, Automated detection and diameter estimation for mesenteric artery using semantic segmentation, bioRxiv, 2020.


Posted on: 11th December 2020


Read preprint (No Ratings Yet)

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here

preLists in the developmental biology category:

EMBL Conference: From functional genomics to systems biology

Preprints presented at the virtual EMBL conference "from functional genomics and systems biology", 16-19 November 2020


List by Jesus Victorino

Single Cell Biology 2020

A list of preprints mentioned at the Wellcome Genome Campus Single Cell Biology 2020 meeting.


List by Alex Eve

Society for Developmental Biology 79th Annual Meeting

Preprints at SDB 2020


List by Irepan Salvador-Martinez, Martin Estermann

FENS 2020

A collection of preprints presented during the virtual meeting of the Federation of European Neuroscience Societies (FENS) in 2020


List by Ana Dorrego-Rivas

Planar Cell Polarity – PCP

This preList contains preprints about the latest findings on Planar Cell Polarity (PCP) in various model organisms at the molecular, cellular and tissue levels.


List by Ana Dorrego-Rivas

Cell Polarity

Recent research from the field of cell polarity is summarized in this list of preprints. It comprises of studies focusing on various forms of cell polarity ranging from epithelial polarity, planar cell polarity to front-to-rear polarity.


List by Yamini Ravichandran

TAGC 2020

Preprints recently presented at the virtual Allied Genetics Conference, April 22-26, 2020. #TAGC20


List by Maiko Kitaoka, Madhuja Samaddar, Miguel V. Almeida, Sejal Davla, Jennifer Ann Black, Dey Lab

3D Gastruloids

A curated list of preprints related to Gastruloids (in vitro models of early development obtained by 3D aggregation of embryonic cells). Preprint missing? Don't hesitate to let us know.


List by Paul Gerald L. Sanchez and Stefano Vianello

ASCB EMBO Annual Meeting 2019

A collection of preprints presented at the 2019 ASCB EMBO Meeting in Washington, DC (December 7-11)


List by Madhuja Samaddar, Ramona Jühlen, Amanda Haage, Laura McCormick, Maiko Kitaoka

EDBC Alicante 2019

Preprints presented at the European Developmental Biology Congress (EDBC) in Alicante, October 23-26 2019.


List by Sergio Menchero, Jesus Victorino, Teresa Rayon, Irepan Salvador-Martinez

EMBL Seeing is Believing – Imaging the Molecular Processes of Life

Preprints discussed at the 2019 edition of Seeing is Believing, at EMBL Heidelberg from the 9th-12th October 2019


List by Dey Lab

SDB 78th Annual Meeting 2019

A curation of the preprints presented at the SDB meeting in Boston, July 26-30 2019. The preList will be updated throughout the duration of the meeting.


List by Alex Eve

Lung Disease and Regeneration

This preprint list compiles highlights from the field of lung biology.


List by Rob Hynds

Young Embryologist Network Conference 2019

Preprints presented at the Young Embryologist Network 2019 conference, 13 May, The Francis Crick Institute, London


List by Alex Eve

Pattern formation during development

The aim of this preList is to integrate results about the mechanisms that govern patterning during development, from genes implicated in the processes to theoritical models of pattern formation in nature.


List by Alexa Sadier

BSCB/BSDB Annual Meeting 2019

Preprints presented at the BSCB/BSDB Annual Meeting 2019


List by Dey Lab

Zebrafish immunology

A compilation of cutting-edge research that uses the zebrafish as a model system to elucidate novel immunological mechanisms in health and disease.


List by Shikha Nayar