Introduction

• X-rays, different layers, 3D

water molecules relay on magnetic field —> low energy absobs radio waves —> release energy again

X-ray, radioactive sugar, combined with CT

• Structural information: X-ray (2D), CT, MRI, …

• Temporal/Dynamic information: Ultrasound, dynamic CT, MRI

• Physiologic metabolic information: perfusion imaging, flow imaging, PET, …

• Microscopic information: OCT, Micro CR

Algorithms for automatic interpretation of biomedical images_ detect, segment, reconstruct, enhance, predict, analyze,…

• Increasing num of images/modalities but not enough doctors to analyse

• Increase the throughput

• At each pixel/voxel assign a new value or a set of values

• New value represents: noise free intensities, higher resolution, artifact free intesities,…

• Information at each pixel: intersity around and at different modalities

• Enhancement algorithm: J(X)=G(f(x))

• At each pixel assing label

• Labels can be: organs, lesions, etc.

• L(x)=S(f(x))

• Segmentation techniques overview: Clustering (K-means), Graph partitioning, Region growing, PDE-based, DIscriminative modeling based (supervised, DNNs) , Generative modeling based (Expectation max)

• Determine spatial transformation between 2 images I(x) und J(x)

• J(T(x)) = I(x) —> map the images

• T can be linear or non-linear

• Techniques: Problem-based, transformation based, non-linear, ladmark based, dense field models, discrete optimization,…

^- Mappping from image to label

• Image size: number of pixels

• Pixel size: the size the pixel occupies in the real worls

• Field of view (FOV): the size the entire image occupies in the real world: Image size * Pixel size

• Standard for storing and transmitting data

• Mostly common used format in practice

• There are also some volumetric formats like NIFT, ITK snap