NFDI4BIOIMAGE
Published September 15, 2025 | Version v1
Image Open

NFDI4BIOIMAGE Calendar August 2025

Creators

  • 1. Data-Driven Imaging in Medicine, University of Applied Sciences and Arts, Göttingen, Germany

Contributors

Data curator:

  • 1. ROR icon Heinrich Heine University Düsseldorf
  • 2. ROR icon NFDI4BIOIMAGE

Description

Image from the NFDI4BIOIMAGE Calendar August 2025.

This image illustrates tissue oxygen saturation in the hand, calculated using various computer-assisted methods and based on hyperspectral and multispectral imaging. The purpose of this image is to compare the perfusion parameters (3 and 4) obtained with multispectral cameras delivering relatively less spectral information but capable of real-time imaging against the perfusion parameters (2) obtained with a hyperspectral medical system delivering large spectral information but not capable of real-time imaging. The picture shows that deep learning methods (4) perform better than classical methods (3) that are not based on artificial intelligence. It lays the groundwork for future real-time quantitative assessment of perfusion during organ transplantation surgeries.

Image Metadata (using REMBI template):

Study

Study description

Quantification of tissue reperfusion using real-time spectral imaging and deep learning

Study type

Study on volunteers

Study Component

Imaging method

(1) RGB imaging

(2) Hyperspectral imaging

(3) and (4) Multispectral imaging

Image component description

(1) RGB image of the hand under normal perfusion.

(2) Perfusion parameter map computed based on hyperspectral imaging (100 spectral bands between 500 and 1000 nm). High perfusion values are represented in red, low perfusion values in blue.

(3) Perfusion parameter map computed based on multispectral imaging (31 spectral bands between 460 and 850 nm) and using the spectral bands that are available but that are less than in (2). Therefore, the result in (3) looks very different from the result in (2).

(4) Perfusion parameter map computed based on multispectral imaging and using a deep neural network. The result in (4) looks similar to the result in (2).

Biosample

Biological entity

Hand

Organism

Homo sapiens

Image data

Image resolution hyperspectral imaging

Spatial Resolution: 480*640 pixels

Spectral Resolution: 500 nm-1000 nm, 100 bands, 5 nm

Image resolution multispectral imaging

Spatial Resolution: 1088*2048 pixels

Spectral Resolution: 16 bands in 460-600 nm; 15 bands in 600-850 nm

Image mode

Reflectance

Submitted via NFDI4BIOIMAGE

Files

image_august.PNG

Files (661.4 kB)

Name Size Download all
md5:402e6ffaae122d7bab385c1a847707ec
661.4 kB Preview Download