Article
Towards automated brain tumour detection with optical coherence tomography
Auf dem Weg zu einer automatisierten Erkennung von Hirntumorgewebe mittels optischer Kohärenztomographie
Search Medline for
Authors
-
Paul Strenge
- Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Deutschland
-
Birgit Lange
- Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Deutschland
-
Veit Danicke
- Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Deutschland
-
Dirk Theisen-Kunde
- Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Deutschland
-
Christin Grill
- Universität zu Lübeck, Institut für Biomedizinische Optik, Lübeck, Deutschland
-
Wolfgang Draxinger
- Universität zu Lübeck, Institut für Biomedizinische Optik, Lübeck, Deutschland
-
Robert Huber
- Universität zu Lübeck, Institut für Biomedizinische Optik, Lübeck, Deutschland
-
Elisa Ducho
- Universitätsklinikum Schleswig-Holstein, Abteilung für Neurochirurgie, Lübeck, Deutschland
-
Matteo Mario Bonsanto
- Universitätsklinikum Schleswig-Holstein, Abteilung für Neurochirurgie, Lübeck, Deutschland
-
Heinz Handels
- Universität zu Lübeck, Institut für Medizinische Informatik, Lübeck, Deutschland
-
Christian Hagel
- Universitätsklinikum Hamburg-Eppendorf, Institute for Neuropathology, Hamburg, Deutschland
-
Ralf Brinkmann
- Medizinisches Laserzentrum Lübeck GmbH, Lübeck, Deutschland; Universität zu Lübeck, Institut für Biomedizinische Optik, Lübeck, Deutschland
| Published: | June 26, 2020 |
|---|
Outline
Text
Objective: Optical coherence tomography (OCT) is a non-invasive technique that can capture 3-dimensional images with micrometer resolution. Prior studies of several groups have shown that OCT has the potential to detect glial tumors by evaluating changes in tissue structure and/or optical properties compared to healthy brain tissue. The aim of our project is the in vivo identification of brain tumours with OCT as a guidance for the surgeon to increase the neurosurgical tumor resection efficiency. The basis for the realization of this goal are segmented datasets, which enable the training of an AI-based classification of tissue.
Methods: Three different OCT systems were used for imaging human glial tumors in vivo (830nm spectral domain (SD) OCT integrated into a surgical microscope) and ex vivo (940nm SD-OCT and 1310nm swept-source MHz-OCT using a Fourier domain mode locked (FDML) laser). Overall, more than 140 human brain samples with different infiltration grades were taken from 20 patients diagnosed with glioblastoma multiforme or other malign brain tumors. For the ex vivo image acquisition, the brain tissue was embedded in a negative agar cuboid. This step simplifies creating H&E stained histological sections in the same orientation as the OCT scans. From every sample, several sections were segmented by a neuropathologist.
Transferring the information of the histological sections to the OCT data set was possible by using structures visible in both image modalities (e.g. blood vessels, hemorrhages) for registration. The segmented OCT data set gained from this process was used to evaluate supervised classification algorithms and start training of neural networks. For tissue discrimination, optical properties such as the attenuation coefficient, but also texture features related to tissue structure were considered.
Results: First examples of the clinical study show that a registration of histological and OCT images was possible to define the ground truth of the tumor segmentation in the OCT images. Furthermore, a spatially resolved representation of the attenuation coefficient provides a good image contrast and confirm that white matter shows a higher signal and more homogeneous signal structure than tumor tissue.
Conclusion: Intraoperative high speed OCT has the potential for in situ tissue monitoring analysis and detection of residual tumor. Thus, intraoperative OCT with real-time data evaluation by AI algorithms may be used for guidance of the neurosurgical resection.
