11-Jun-2019 - Rheinische Friedrich-Wilhelms-Universität Bonn

How Artificial Intelligence Detects Rare Diseases

Every year, around half a million children worldwide are born with a rare hereditary disease. Obtaining a definitive diagnosis can be difficult and time consuming. In a study of 679 patients with 105 different rare diseases, scientists from the University of Bonn and the Charité - Universitätsmedizin Berlin have shown that artificial intelligence can be used to diagnose rare diseases more efficiently and reliably. A neural network automatically combines portrait photos with genetic and patient data.

Many patients with rare diseases go through lengthy trials and tribulations until they are correctly diagnosed. "This results in a loss of valuable time that is actually needed for early therapy in order to avert progressive damage," explains Prof. Dr. med. Dipl. Phys. Peter Krawitz from the Institute for Genomic Statistics and Bioinformatics at the University Hospital Bonn (UKB). Together with an international team of researchers, he demonstrates how artificial intelligence can be used to make comparatively quick and reliable diagnoses in facial analysis.

The researchers used data of 679 patients with 105 different diseases caused by the change in a single gene. These include, for example, mucopolysaccharidosis (MPS), which leads to bone deformation, learning difficulties and stunted growth. Mabry syndrome also results in intellectual disability. All these diseases have in common that the facial features of those affected show abnormalities. This is particularly characteristic, for example, of Kabuki syndrome, which is reminiscent of the make-up of a traditional Japanese form of theatre. The eyebrows are arched, the eye-distance is wide and the spaces between the eyelids are long.

The used software can automatically detect these characteristic features from a photo. Together with the clinical symptoms of the patients and genetic data, it is possible to calculate with high accuracy which disease is most likely to be involved. The AI and digital health company FDNA has developed the neural network DeepGestalt, which the researchers use as a tool of artificial intelligence for their study. "PEDIA is a unique example of next-generation phenotyping technologies," said Dekel Gelbman, CEO of FDNA. "Integrating an advanced AI and facial analysis framework such as DeepGestalt into the variant analysis workflow will result in a new paradigm for superior genetic testing".

Researchers train the neural network with 30,000 images

The scientists trained this computer program with around 30,000 portrait pictures of people affected by rare syndromal diseases. "In combination with facial analysis, it is possible to filter out the decisive genetic factors and prioritize genes," says Krawitz. "Merging data in the neuronal network reduces data analysis time and leads to a higher rate of diagnosis.”

The head of the Institute of Genomic Statistics and Bioinformatics at the UKB has been working with FDNA for some time. "This is of great scientific interest to us and also enables us to find a cause in some unsolved cases," said Krawitz. Many patients are currently still looking for an explanation for their symptoms.

The study is a team effort between computer science and medicine. This can also be seen in the shared first authorship of the computer scientist Tzung-Chien Hsieh, doctoral student at the institute of Professor Krawitz, and Dr. Martin Atta Mensah, physician at the Institute of Medical Genetics and Human Genetics of the Charité and Fellow of the Clinician Scientist Program of the Charité and Berlin Institute of Health (BIH). Prof. Dr. Stefan Mundlos, Director of the Institute of Medical Genetics and Human Genetics at the Charité, also participated in the study, as did over 90 other scientists.

"Patients want a prompt and accurate diagnosis. Artificial intelligence supports physicians and scientists in shortening the journey," says Dr. Christine Mundlos, Deputy Managing Director of the alliance of patients with chronic rare diseases (ACHSE) e.V. "This also improves the quality of life of those affected to some extent."

  • Tzung-Chien Hsieh MS et al.; "PEDIA: Priorization of Exome Data by Image Analysis, Genetics in Medicine"; Genetics in Medicine; 2019

Rheinische Friedrich-Wilhelms-Universität Bonn

Request information now

Recommend news PDF version / Print

Share on

Facts, background information, dossiers
  • hereditary diseases
  • diagnostics
  • rare diseases
  • artificial intelligence
  • neural networks
More about Universität Bonn
  • News

    How cells recognize uninvited guests

    Until now, the immune sensor TLR8 has remained in the shadows of science. A research team led by the University of Bonn has now discovered how this sensor plays an important role in defending human cells against intruders. The enzymes RNaseT2 and RNase2 cut ribonucleic acids (RNAs) of bacte ... more

    New method identifies aggressive breast cancer

    Aggressive forms of breast cancer often manipulate the immune response in their favor. This manipulation is revealed in humans by the same immunological "signature" as in mice. This is shown by a study carried out by scientists from the University of Bonn together with Dutch colleagues. The ... more

    Shipment tracking for "fat parcels" in the body

    Without fat, nothing works in the body: These substances serve as energy suppliers and important building blocks - including for the envelopes of living cells. Numerous diseases are related to disorders in the fat metabolism, such as obesity or cancer. Researchers from the LIMES Institute a ... more

More about Charité
  • News

    Could the blood of COVID-19 patients be used to predict disease progression?

    Researchers from Charité – Universitätsmedizin Berlin and the Francis Crick Institute have identified 27 proteins which are present at different levels in the blood of COVID-19 patients, depending on the severity of their symptoms. These biomarker profiles could be used to predict disease p ... more

    Diagnostics, meet CRISPR

    A new diagnostic test to quickly and easily monitor kidney transplant patients for infection and rejection relies on a simple urine sample and a powerful partner: the gene-editing technology CRISPR. Michael Kaminski, who developed it, leads a new Emmy Noether Group at the MDC & Charité. The ... more

    Recognise and control new variants of the deadly Ebola virus more quickly

    The situation is extraordinary: there have only ever been four declarations of public health emergencies of international concern in the past and now there are two at the same time. Whilst the risks associated with the novel coronavirus are still unclear, people in the Democratic Republic o ... more