Training and Verification of Artificial Neural Networks for the Detection of Anatomical Landmarks on the Liver Surface
The aim of this Bachelor thesis is to investigate the use of artificial neural networks (ANNs) for the detection of anatomical landmarks on the liver surface in laparoscopic video images. The goal is to develop an algorithm that can identify these landmarks accurately and quickly, which is an important part of an intraoperative navigation tool in laparoscopic liver surgery. The first part of the project will involve training ANNs using a dataset of laparoscopic images of the liver. The images are annotated by a medical student with the location of various anatomical landmarks, such as the falciform ligament or the anterior ridge. Different architectures and hyperparameters of ANNs will be explored to achieve the best accuracy and efficiency. The effectiveness of the developed models will be evaluated using various metrics such as precision, recall, F1 score, and receiver operating characteristic (ROC) curves. In the second part of the project, the developed models will be verified using a separate dataset of laparoscopic images. The accuracy and efficiency of the developed algorithm will be evaluated using statistical analysis and visual inspection of the results. The main contributions of this thesis are the development and evaluation of ANNs for the detection of anatomical landmarks on the liver surface in laparoscopic video images. In future, the detected anatomical landmarks can be used for the alignment with the preoperative 3D CT-model. The project will demonstrate the potential of ANNs in medical image analysis and the challenges associated with training and verifying these models.