Introduction

Fiber based common path optical coherence tomography (CP-OCT) may find its application as a tissue sensitive guidance tool in pituitary adenoma resection. In the envisioned application, the pituitary adenoma is aspirated and characterized in real-time within the suction tool using (one dimensional) CP-OCT A-scans. To obtain a robust and reliable machine learning (ML) model for tissue discrimination, we recorded CP-OCT data on healthy pituitary gland as well as adenoma tissue embedded in paraffin. In this report, we present the results of thirty paraffin-embedded tissue samples.

Methods

To obtain enough training material for the data model, a two-dimensional matrix of equidistant 100 x 100 OCT A-scans was recorded for each paraffin-embedded tissue sample. Care was taken to set the minimum spacing between A-scans significantly higher than the spatial resolution of the applied CP-OCT device, to record different sample positions. For every A-scan, the following parameters were calculated: attenuation coefficient, residuum, intensity distribution of the backscattered light, and spectral properties. After preprocessing and scaling steps, various clustering methods were applied on these parameters to categorize the A-scans on different sample spots.

Results

En-face images based on different extracted A-scan parameters qualitatively show the varying ability of the parameters to discriminate between sample regions. The clustering methods detect multiple categories of sample areas. Blood within the tissue samples may lead to masking of tissue differences.

Conclusion

The data of thirty paraffin-embedded tissue samples show promising results in discriminating adenoma from healthy pituitary gland tissue by using in vivo OCT A-scans. The availability of haematoxylin-eosin-stained histological thin sections of the same samples later in the study will allow the creation of more accurate supervised machine learning models for the targeted application of guided pituitary adenoma resection.

In head and neck surgery, continuous intraoperative tissue differentiation is of great importance to avoid injury to sensitive structures such as nerves and vessels. Hyperspectral imaging (HSI) with neural network analysis could support the surgeon in tissue differentiation.

A 3D Convolutional Neural Network with hyperspectral data in the range of 400-1000 nm is used in this work. The acquisition system consisted of two multispectral snapshot cameras creating a stereo-HSI-system. For the analysis, 27 images with annotations of glandular tissue, nerve, muscle, skin and vein in 18 patients undergoing parotidectomy are included. Three patients are removed for evaluation following the leave-one-subject-out principle. The remaining images are used for training, with the data randomly divided into a training group and a validation group.

In the validation, an overall accuracy of 98.7% is achieved, indicating robust training. In the evaluation on the excluded patients, an overall accuracy of 83.4% has been achieved showing good detection and identification abilities.

The results clearly show that it is possible to achieve robust intraoperative tissue differentiation using hyperspectral imaging. Especially the high sensitivity in parotid or nerve tissue is of clinical importance. It is interesting to note that vein was often confused with muscle. This requires further analysis and shows that a very good and comprehensive data basis is essential. This is a major challenge, especially in surgery.

A modern area of research in cancer treatment is magnetic drug targeting (MDT) with superparamagnetic iron oxide nanoparticles (SPIONs). In order to understand the processes involved in MDT in more detail and to be able to perform this therapy as efficiently as possible, a monitoring system for the spatial distribution of SPIONs in biological tissue is required. One approach is to use magnetomotive ultrasound (MMUS) to monitor the spatial distribution over time. However, the spatial distribution of SPIONs cannot be quantitatively determined applying basic MMUS algorithms. Therefore, MMUS has been extended by a simulation part to quantitatively determine the spatial distribution of SPIONs. This extended MMUS algorithm requires the material parameters and the geometry of the target tumorous tissue. In this contribution, we describe an ultrasound-based eigenfrequency analysis combined with an inverse simulation-based method to determine the mechanical parameters shear and Young's modulus of a tissue mimicking phantom. The presented approach yields a good estimate of the Young's modulus compared to the result from a compression test.

Abstract wird bis 24.4. nachgereicht. Absprache zw. Prof. Pott und Prof. Grabow.