Jump to content
Powered by

Database instead of a book – software facilitates diagnosis of liver diseases

The bachelor’s theses of Matthias Hillert and Pascal Laube, students at the Konstanz University of Applied Sciences (HTWG), involved the development of a computer programme that facilitates the analysis of computed tomography images of the liver by comparing them to similar images stored in a database. It is a tool that has the potential to be used in other areas, such as the identification of bone tumours.

The bachelor’s theses of the students Matthias Hillert (centre) and Pascal Laube (right) from Konstanz were supervised by Prof. Dr. Christian Johner (left). © HTWG Konstanz

Computed tomography (CT) is a common method used to examine the liver, in particular for the diagnosis of lesions and tumours. CT involves collecting X-ray images of a patient’s body from multiple directions. A computer creates separate images, called slices, of the liver or other organs. At present, clinical staff tend to analyse the slices by comparing them to images in books. “This is a time-consuming procedure and is prone to errors,” said Dr. Christian Johner, professor and lecturer in medical informatics at the Konstanz University of Applied Sciences and director of the Institute for Healthcare IT. Together with two of his students, Matthias Hillert and Pascal Laube, he took on the challenge of developing computer software for analysing CT images.

The goal was to develop a programme that could compare new CT images with CT images stored in a database that already linked to diagnoses. “Although the software does not expand the number of possibilities of detecting tumours at an early stage, the automated comparison of images with a large number of similar database images would nevertheless enable quick and reliable diagnoses,” said Hillert and Laube. The students then had to develop algorithms that would enable the comparison of CT images, and this was not an easy task. Hillert and Laube combined two algorithms, one of which compares the shape and geometry of the areas under investigation, while the other compares the texture and structure of the areas. The students also developed solutions for dealing with artifacts, such as those that arise when patients move during the imaging process. These solutions involved the integration of specific filters into the software. The treating doctor can then see images that have the greatest degree of similarity to those that he/she has obtained, along with information on the diagnoses made. The doctor can also feed the databank with his/her own images and associated diagnoses. “Our algorithms are still very general, but it is theoretically possible to adapt them to new developments in CT diagnostics,” Hillert and Laube explained. 

The images need to be specifically processed before they can be analysed: the images in the top row show the extraction of a liver section from a more comprehensive image; the photos in the bottom row show the filtering of an image with high background noise.
The images need to be specifically processed before they can be analysed: the images in the top row show the extraction of a liver section from a more comprehensive image; the photos in the bottom row show the filtering of an image with high background noise. © private

Potential to extend the software to bone tumours and head area

Professor Christian Johner has been doing research at the interface of medicine and informatics for quite some time. However, he has not been the only person to supervise the two students. Other supervisors included experts with practical experience such as Matthias Franz, professor at the Institute for Optical Systems at the Konstanz University of Applied Sciences, and Dr. Peter Köhler, a medical specialist and partner in the Prof. H. Zwicker & Partner Clinic, a clinic that specialises in diagnostic radiology, radiotherapy and nuclear medicine. Köhler drew the students’ attention to the liver, as the liver is the one organ where it is particularly difficult to distinguish between different diseases (e.g., different types of jaundice). CT scans provide more detailed information than standard X-rays, and hence more information related to diseases of the liver. “In addition, the algorithms are relatively effective in identifying the abdominal organs, which makes it easier for us to compare the images,” Hillert and Laube explained. The software can also be applied to images of the head area, where the likelihood of confusion between diseases is also quite high. The students also believe that the software can be adjusted to the diagnosis of bone tumours.

Hillert and Laube have received numerous prizes for their achievements, including the 3rd prize of the Bodensee Innovationspreis (Lake Constance Innovation Prize) sponsored by the Lienhard Office Group, and the Karl Steinbuch Scholarship awarded by MFG Foundation Baden-Württemberg. “We are using the grant to adapt the software so that it can be effectively used by radiologists,” Hillert and Laube said, adding that they are also supported by the Freiburg University Medical Centre in this project. They can envisage working with companies. “We believe that our software has the potential to be used and sold as plugin to manufacturers of clinical information systems,” Johner explained. However, before this can happen, Hillert and Laube need to finish their master’s degree course at the Konstanz University of Applied Sciences.

Further information:
Institute for Healthcare IT 
Prof. Dr. Christian Johner
Villa Rheinburg
Reichenaustr. 1
78467 Konstanz
E-mail: christian.johner(at)johner-institut.de

Website address: https://www.gesundheitsindustrie-bw.de/en/article/news/database-instead-of-a-book-software-facilitates-diagnosis-of-liver-diseases