Powered by
06:47 CET
RC 105 - Everything you need to know about 3D post-processing
Imaging Informatics Artificial Intelligence & Machine Learning
Wednesday, February 28, 08:30 - 10:00
Room: M 2
Moderator: E. Sorantin (Graz/AT)

Chairperson's introduction
E. Sorantin; Graz/AT
Learning Objectives

1. To learn about the state of the art in 3D post-processing.
2. To understand how 3D post-processing can most optimally be used in daily clinical practice.
3. To appreciate how automated 3D post-processing and quantification will lead to increased use of 3D visualisations for diagnostics and therapy planning, over 2D viewing.


Progress in imaging technology equipment enables scanning patients in high geometrical and temporal resolution as well as in multidimensional space (e.g. 4D). The amount of resulting data cannot be read any more in 2D as done in the last millennium. Furthermore, advances in computational power enable the use of sophisticated processing algorithms in real time. Thus reading in 2D, as done in the previous millennium, will be gradually replaced by volumetric reading as well as extracting diagnostic information from parametric images. Moreover, for personalized medicine, radiology has to deliver more detailed information, especially to measure tumour volumes or characterize contrast uptake on perfusion imaging. To get familiar with those now really emerging techniques, three well-known speakers will cover essential subtopics and provide a road map on how to migrate from the reading style in the last millennium to that in the current millennium.

A. 3D post-processing in 2018
A. Alberich-Bayarri; Valencia/ES
Learning Objectives

1. To learn about recent advances in 3D post-processing techniques.
2. To understand how these techniques can be used in clinical practice now.
3. To learn new tips and tricks to use in your daily practice.


One of the most important developments in radiological interpretation is the need for the incorporation of advanced tools to assist the specialist in the study evaluation. Automated segmentation of structures based on convolutional neural networks (CNN) in the frame of deep learning would allow to significantly increase the efficiency of the study evaluation by the radiologist. Although the detailed segmentation of organs is still intricate in the field of abdomen and modalities like MR, current technology allows for the automated detection of the organs' location and identification of most of the tissue using bounding boxes. These applications may be used today in clinics for the automated assessment of tissue properties. A clear example is the automated detection and identification of vertebrae centroids, which allows for the acceleration of the radiologist reading process in spine CT examinations while it also allows for the automated calculation of trabecular bone quality properties in each identified vertebrae, therefore providing a high value to perform osteoporosis population studies without the need for a user interaction. These algorithms have been recently labelled as zero-click solutions and will provide a paradigm shift in 3D post-processing for radiologists, having the results of the 3D assessment already generated in their PACS even before starting review of the study.

B. Making better use of your 3D package: tips and tricks
P. M. A. van Ooijen; Groningen/NL
Learning Objectives

1. To learn about the functionality of state-of-the-art 3D packages.
2. To understand the pitfalls in use of 3D post-processing.
3. To appreciate the need for training in 3D post-processing techniques.


Advanced visualization, simulation and planning software is increasingly used in clinical practice providing a shift from 2D to 3D visualization, processing and interpretation. With this ongoing trend the radiological profession should not only focus on the diagnosis to be made, but also on the utilization of our imaging data in patient simulation, planning, and treatment. Current functionality moves in this direction with providing extensive possibilities for support of surgical interventions and treatment planning in 3D including the advent of Virtual and Augmented Reality. With this 3D is also moving into the operating theater. Although these new possibilities are interesting and exiting one should be very aware of the pitfalls that come with 3D visualization and processing of data. This not only includes the technical but also the procedural pitfalls where image acquisition optimal for diagnosis is not always optimized for the intended use by the referring physician. To adequately use the new techniques and to provide optimal support from radiology to the referring physicians training is required and dedicated staff should be involved in this process.

C. Interpretation of 3D processing results: from image to volume reading
T. Frauenfelder; Zurich/CH
Learning Objectives

1. To learn about different developments in creating 3D anatomical and functional models for diagnostic and therapy planning purposes.
2. To understand the pros and cons of such technologies.
3. To appreciate that automated 3D image analysis will lead to new ways in which diagnosis and therapy planning will be performed.


The widespread introduction of multidetector computed tomography (MDCT) has revolutionized the field of computed tomography (CT). This revolution can be attributed to three primary properties of MDCT: its ability to produce a vast quantity of volumetric data in a reduced amount of time, the high resolution, and the ability to create isotropic voxel data and, consequently, reliable multiplanar and three-dimensional (3D) reconstructions. Diagnostic approaches that rely solely on axial reconstructions of MDCT data are often insufficient for formulating an accurate diagnosis or for documentation of clinical cases. Specialized 3D reconstruction techniques permit the visualization of anatomical details, which would be difficult to evaluate using axial reconstructions alone. Such details may require the use of oblique or curved reconstructions, or more complex methods, such as maximum intensity projection (MIP), minimum intensity projection (MinIP), surface-shaded volume rending (SS-VRT), and virtual endoscopy. For example, small pulmonary nodules can only be rapidly and reliably identified through the use of MIP Slab slices. The current trend is to merge the routine diagnostic console and 3D reconstruction workstation. The integration of 3D reconstruction utilities into the standard bi-dimensional diagnostic software has increased the number of operations possible on each exam data, greatly increasing the perceived complexity of CT diagnosis. Although many of us believe that the use of 3D reconstructions greatly increases total exam evaluation time, there are reports show how using 3D reconstruction techniques for examining volumetric data are effective and also improve the speed of interpretation, recognition, and description of specific clinical conditions. Many of these reconstruction techniques are of particular importance for the analysis of subspecialty exams, as for example the 3D depiction and quantification of lung emphysema.

Panel discussion: Will we still look at 2D images in 10 years' time?

Image interpretation of 3D results: from image reading to volume reading.

This website uses cookies. Learn more