Von wandernden Ionen über das Zellgewebe bis hin zur Aufzeichnung eines EKG: Die Softwaresimulation des menschlichen Herzens ermöglicht maßgeschneiderte Therapien. Am Karlsruher Institut für Technologie (KIT) haben Forscher ein Computermodell des menschlichen Herzens entwickelt. Die Wissenschaftler sind mittlerweile sogar schon so weit, dass sie das Modell auf die individuellen Eigenschaften eines einzelnen Patienten maßschneidern können. Diese Simulation hat einen erheblichen Nutzen für die medizinische Praxis.
To measure blood flow distributions within the lungs at bedside, Electrical Impedance Tomography measurements based on conductive indicator signals have been recently proposed. The first passage of the indicator signal through the lungs is exploited, but needs to be separated from a superimposed slow drift signal. Two fitting approaches are presented in this paper to accomplish this separation task. The accuracy of estimated first pass signal features is investigated on a synthetic data base. Both algorithms alter the shape of the indicator signal similarly. The algorithms are finally tested on real data from a preclinical porcine study.
A. Naber, D. Berwanger, G. K. Steinberg, and W. Nahm. Spatial gradient based segmentation of vessels and quantitative measurement of the inner diameter and wall thickness from ICG fluorescence angiographies. In SPIE Photonics West, vol. 11229 1122916-2, 2020
During neurovascular surgery the vascular function can be checked intraoperatively and qualitatively by observing the blood dynamics inside the vessel via Indocyanine Green (ICG) Fluorescence Angiography. This state-of-the- art method provides the surgeon with valuable semi-quantitative information but needs to be improved towards a quantitative assessment of vascular volume flow. The precise measurement of volume flow rely on the assumption that both the inner geometry of the blood vessel and the blood flow velocity can be precisely obtained from Fluorescence Angiography. The correct reconstruction of the inner diameter of the vessel is essential in order to minimize the propagated error in the flow calculation. Although ICG binds specifically on blood plasma proteins the fluorescence light radiates also from outside the inner vessel volume due to multiple scattering in the vessel wall, causing a fading edge intensity contrast. A spatial gradient based segmentation method is proposed to reliably estimate the inner diameter of cerebral vessels from intraoperative Fluorescence Angiography images. As result the minimum of the second deviation of the intensity values perpendicular to the vessels edge was identified as the best feature to assess the inner diameter of artificial vessel phantoms. This method has been applied to cerebrovascular vessel images and the results, since no ground truth is available, comply with literature values.
Confocal laser endomicroscopy (CLE) has found an increasing number of applications in clinical and pre-clinical studies, for it allows intraoperative in-situ tissue morphology at cellular resolution. CLE is considered as one of the most promising systems for in-vivo pathological diagnostics. Miniaturized imaging probes are designed for intraoperative applications. Due to less sophisticated optical design, CLE systems are more prone to image aberrations and distortions. While diagnostics with CLE takes reference from the corresponding histological images, the determination of the resolution and aberrations of the CLE systems becomes essential. Thereby on-site quality check of system performance is required. Additionally, these compact systems enable a field of view of less than half square millimeter without zooming function, which makes it difficult to correlate human vision to the microscopic scenes. Therefore, it is necessary to have defined microstructures working as a test target for CLE systems. We have extended the 2D bar pattern in 1951 USAF test chart to 3D structures for both lateral and axial resolution assessment, since axial resolution represents the optical sectioning ability of CLE systems and is one of the key parameters to be assessed. The test target was produced by direct laser writing. Yellow-green fluorescence emission can be excited at 488 nm. It can also be used for other fluorescence microscopic imaging modalities in the corresponding wavelength range.
Student Theses (1)
K. Hii. Reorientation of an Atrial Model to Simulate 12-lead ECG Signals: An Overfitting and Data Augmentation Problem. Institute of Biomedical Engineering, Karlsruhe Institute of Technology (KIT). Bachelorarbeit. 2020
The present study of different atrial flutter mechanisms remains a very complex subject. Without the use of invasive mapping techniques or just by observing the 12-lead ECG signals, it is impossible to differentiate between atrial flutter mechanisms. A more sophisticated approach like a radial basis neural network classifier is implemented in this thesis to classify atrial flutter signals according to its mechanism. However, in order to have a good classifier, two important aspects need to be considered: a huge amount of data which, at the same time, does not cause overfitting. Data from previous studies were used as a benchmark to assess the performance of the classifier by enlarging the available dataset. One way to feed the classifier new data is by using data augmentation methods. In our study, we simulated different rotations around all 3 axes of the atrial model to generate new 12-lead ECG signals. We also investigated the potential problem of overfitting in the process. We started by first doing a correlation analysis of the ECG signals to have an idea how much signals could change at each rotation. Here, the signals between the initial position and each of the rotated position were compared. We found out that within a range of ±10◦, there were, in most cases, correlation coefficients higher than 0.75 which might not be useful for machine learning applications. We implemented different scenarios to investigate which train and test dataset division would improve the classifier accuracy or trigger an undesired overfitting problem. We found out that adding rotations to the atrial model as a means of data augmentation improved the performance of the classifier for some mechanisms. This, however, was valid if a part of the atrial model dataset was also used for training. From this, we learned that we could achieve an individual increase of 12 - 25% in accuracy using atrial models whose data was partially used in the train set. The initial idea was to train the classifier with atria models and test with ’unseen’ atria models. Yet, we noticed that the classifier did not perform as well on unknown atrial models as accuracies observed were lower than the benchmark. To achieve a higher accuracy, we concluded that augmenting the dataset of only 2 atrial models were not enough to improve the overall classifier accuracy and more atrial geometries were needed to investigate the possible improvements they could bring.