.  Home  .  Lehre  .  Studentische Arbeiten  .  Masterarbeiten  .  Ausschreibung

Masterarbeit: Deep-learning-based analysis of hand radiographs to detect deviations between skeletal and biological maturity

The skeletal maturity of human beings can be observed by a series of specific phases. These phases are particularly seen in the wrists and hands. Bone age screening is frequently performed in a pediatric context to evaluate the growth of the patient and to detect endocrine disorder and pediatric syndromes as soon as possible. It is common practice that the detection of the bone age and the following comparison with the chronological age is based on visual evaluation of the the hand and the wrist. Radiologists compare the radiographs of their patients with reference images to detect unusual deviations. This method is time consuming, expensive and highly depends on the expertise and experience of the radiologist. With this thesis multiple approaches will be tested to process radiographs and to compare the performance of different networks and architectures (VGG16, V3, etc.). Special attention must be paid to the preprocessing of the data. Since the original dataset only has a size of less than 2000 annotated images, adequate data augmentation is an indispensable step. Furthermore male and female children develop at different rates. This also affects their bone age and should be included in the preprocessing steps. As a further step, the possibilities of an appropriate presentation of the processed images should be evaluated, since the results have to be verified by an radiologist.

Schwerpunkte / Ideen:

  • Verarbeitung von Röntgen-Bilddaten
  • Konzepte zur automatisierten Verarbeitung großer Datenmengen


Prof. Dr. D. Kranzlmüller

Dauer der Arbeit:

  • Masterarbeiten: 6 Monate

Anzahl Bearbeiter: 1