Frida Holmvik and Evelina Persson will present their master's thesis "Master's Thesis - Investigation of Methods for Automating the Localization of Healthy or Unhealthy Lymph Node Tissue for Magnetomotive Imaging" on Wednesday June 5th at 14:15 in MH:309A. The presentation will be in Swedish.

Rectal cancer is a frequently diagnosed cancer type worldwide, approximately 700,000 people are diagnosed every year. The mortality is high due to the high risk of metastasis where it mainly spreads through the lymphatic drainage pathways. Unfortunately, there are no accurate methods used in the clinic for determining the spread, resulting in that extensive colorectal resection is the primary treatment to ensure that no metastasis is missed. For the early colorectal cancer cases, it has been found that only 10% of the colorectal resections involved lymph node metastasis. To reduce the number of unnecessary tissue resections, NanoEcho AB is developing a device to differentiate between healthy and metastasized lymph node tissue with magnetomotive ultrasound. Magnetomotive ultrasound is a novel method utilizing ultrasound together with a magnetic field and iron oxide-based nanoparticles operating as a contrast agent. The magnetic field causes the magnetic nanoparticles to oscillate and the displacement is detected and visualized by the algorithm. In this thesis, two methods were investigated as potential tools for automatically providing information about the distribution of healthy or unhealthy lymph node tissue, e.g. metastasis, by utilizing a software algorithm in combination with magnetomotive ultrasound. Data from tissue mimicking phantoms and synthetic generated data were used for the investigation. Firstly, the possibility to predict the localization of healthy lymph node tissue using an artificial neural network trained on the available data was investigated. Secondly, a computationally simple metric was investigated, defined as the center of maximum displacement. The result showed that the most promising approach of these two was to utilize an artificial neural network trained on a combination of phantom and synthetic data. With that approach it was possible to automatically distinguish healthy tissue in some of the phantom test data points. However, it was not successful to automatically distinguish healthy tissue using a model trained on only displacement data or synthetic data. With the second approach, the relative shift of center of maximum displacement from the center of the lymph node was compared for lymph nodes, with different positions of unhealthy tissue or solely healthy tissue. In most cases, it could not be stated with certainty that there was a difference leading to the conclusion that a simple metric in this implementation does not fulfil its purpose.