Sub-Project: Near-Infrared real-time Monitoring of the Collateral Network (real-time cnNIRS)
No non-invasive real-time method for monitoring spinal cord tissue perfusion and oxygenation is clinically available (Etz et al. 2013). The use of near-infrared spectroscopy of the paraspinous collateral network (cnNIRS) as an indirect non-invasive method for real-time monitoring during and after extensive aortic aneurysm repair has recently been validated in large animal experiments (von Aspern et al. 2016) and found feasible in small clinical studies (Etz et al. 2013, Luehr et al. 2016). Clinical experience with this method is not available. Our group recently published extensive award-winning experimental data (cnNIRS validation with Laserdoppler-flow) on this modality suggesting that clinical use is both feasible and important.
By using cnNIRS as a non-invasive real-time monitoring modality during and after MISACE, open and endovascular thoracoabdominal aortic aneurysm repair, we aim to evaluate its sensitivity and specifity on a large homogenous cohort in order to detect signs of pending spinal cord ischemia as early as possible during and after these procedures as an adjunct to close clinical examination.
Sub-project: Identification of clinically relevant spinal cord ischaemia (SCI) markers in cerebrospinal fluid (CSF) and development of an easily applicable bed-side test
Organ ischaemia is usually detected in the clinical setting by biomarkers from blood samples such as enzymes or structural proteins, which can be used to confirm diagnoses such as heart attacks. Characteristic enzyme serum levels and established clinical tests are available for most organs and are used for postoperative ischaemia monitoring on a routine basis— except for the spinal cord. This is particularly inauspicious since spinal cord ischaemia (SCI) often times occurs with a delay, i.e. in the early postoperative period after TAAA repair. (Etz et al. 2008a) Biomarkers for the real-time detection of emerging SCI would potentially provide time to intervene (e.g. to elevate systemic blood pressure and/or blood oxygen saturation) and would be of great benefit in preventing devastating complications such as paraplegia, particularly in the early postoperative period after TAAA repair. Detection of SCI biomarkers has not yet been successfully established for the clinic because the levels of specific markers such as S100β, neuron-specific enolase, neurofilament light chain, and glial fibrillary acidic protein neuron specific enolases or heat shock proteins do not reach detectable serum levels and unspecific ischemia markers (e.g. lactate) succumb to excessive dilution. Even though elevated concentrations of neurochemical serum biomarkers have been identified in patients with SCI, these do not provide a sensitive prognostic tool. (Pouw et al. 2009).
Cerebrospinal fluid (CSF) is often used in TAAA repair to reduce CSF pressure and improve spinal cord perfusion, (Coselli et al. 2002), meaning that samples are readily available. The advantage of using CSF to detect ischemia markers arises from minimal dilution, a favourable nerve-tissue-to-CSF ratio and slow fluid exchange rates allowing for rapid accumulation of possible markers. Moreover, the fact little clinical research has been done in this area because of difficult access to CSF means that the chances of success are comparatively good.
The objectives of this sub-project are:
Development of a protocol for routine collection of CSF during the multicentre trial PAPA-ARTiS during and after MISACE (selected centres) and open / endovascular TAAA repair from CSF drainage systems. Validation of possible Biomarkers (N100, HSPs, lactate) of SCI in CSF and correlation of specific marker levels with data from side project #1 (real-time cnNIRS – non-invasive local CN perfusion and oxygenation) and #3 (Pim-PaP – CN perfusion and oxygenation imaging), steps toward development of a clinically applicable bed-side test for SCI from CSF.
Subproject: Patient-based individual modelling of Paraspinal Collateral Network Perfusion after Segmental Artery Occlusion (PimPaP)
The introduction of software systems for intelligent decision-support into the clinical workflow of interventional procedures is primarily a response to the increasing amount of patient-individual data and its consideration in the treatment process. However, the machine-readability of such patient data, as well as the inference of relevant information, is still restricted. The extraction of expert knowledge from the respective medical domains is laborious and often counter-productive to the overall concept of assistance systems. Therefore, the Innovation Center Computer Assisted Surgery (ICCAS) of the University Leipzig is following the approach of digitally modelling the patient as well as the perioperative therapy processes. Using this approach, a software-internal representation of the individual patient state is generated at different time points during the treatment process. Investigating the temporal changes of specific patient data (vital, pathological, and procedure parameters) leads to the abstraction of more generalised representation of the treatment process and, furthermore, enables the comparison of new patients with this treatment representation.
The goal is to establish a general understanding of the coiling patterns and their impact on the convalescence of spinal perfusion and the clinical outcome. Initially, a patient model needs to be developed, taking into consideration multi-modal data, e.g. imaging & procedure parameters as well as cardiovascular and pathology parameters. Following the model definition, an analysis of the MISACE procedure workflow will be conducted, identifying the kay treatment steps and stages involved. From this step, a procedure process model will be developed that represents the generalised workflow for a patient receiving MISACE treatment. Subsequently, the correlation between coiling patterns and patient tolerance can be mapped to the study results. For the support of the treatment planning, a prediction method to estimate patient-individual consequences of coiling patterns on the overall procedure risk, the minimal recuperation time and ultimately, the clinical outcome will be developed and evaluated. A therapy decision-support module for MISACE procedures will then be developed to support the pre-operative treatment planning situation including the prediction of consequences for coiling patterns on the clinical outcome as well as the overall procedure risk.
The patient´s situation is formally described to be computer-readable. For this step, we are using both new taxonomic definitions as well as existing ontological knowledge bases [Uciteli et al. 2011, H. Herre, 2010]. The primary task is the generation of a patient model representation including parameters of the demographic background, imaging data, anamnesis, diagnosis, etc. Here, methods of Machine Learning are employed. The individual patient data gathered from the trial is stored in a knowledge base. Patient-specific parameters are extracted, categorised and stored in an overall patient representation. From the knowledge base potential, treatment-relevant parameter candidates are compiled and surveyed by clinical experts. Subsequently, the MISACE procedure is analysed, and a process model representation developed. For the process model definition, the substantial research results of ICCAS are adopted [Neumuth et al. 2011a, Neumuth et al. 2009, Nuemuth et al. 2011b, Neumuth et al. 2011c, Cypko et al. 2017]. The process model is specified for the MISACE treatment arm and instantiated as individual treatment workflows. In the following generalisation step, a patient-independent treatment process model is created. The development of an estimation method to identify the ideal MISACE stage configuration is based on the highest correlation of parameters between patient models and process models. The outcome prediction system for the MISACE procedure is developed to include patient parameters according to the current treatment workflow step. A proof-of-concept study is conducted to investigate the overall method performance and quality. For the modelling activities, previous ICCAS research results are included and further developed.
von Aspern K, Haunschild J, Hoyer A et al. Non-invasive spinal cord oxygenation monitoring: validating collateral network near-infrared spectroscopy for thoracoabdominal aortic aneurysm repair. Eur J Cardiothorac Surg. 2016.
Coselli JS, LeMaire SA, Köksoy C, Schmittling ZC, Curling PE. Cerebrospinal fluid drainage reduces paraplegia after thoracoabdominal aortic aneurysm repair: results of a randomized clinical trial. J Vasc Surg. 2002 Apr;35(4):631-9.
Cypko MA, Stoehr M, Kozniewski M, Druzdzel M, Dietz A, Lemke HU. Validation workflow for a clinical Bayesian network model in multidisciplinary decision making in head and neck oncology treatment. Int J Comput Assist Radiol Surg [Internet]. 2017 Feb;
Etz CD, von Aspern K, Gudehus S et al. Near-infrared spectroscopy monitoring of the collateral network prior to, during, and after thoracoabdominal aortic repair: a pilot study. Eur J Vasc Endovasc Surg. 2013.
Etz CD, Luehr M, Kari FA, Bodian CA, Smego D, Plestis KA, Griepp RB. J Thorac Cardiovasc Surg. 2008a Feb;135(2):324-30. Paraplegia after extensive thoracic and thoracoabdominal aortic aneurysm repair: does critical spinal cord ischemia occur postoperatively?
H. Herre, “The Ontology of Medical Terminological Systems - Towards the Next Generation of Biomedical Ontologies.” In: R. Poli, L. Obrst (ed.) Theory and Applications of Ontology. Vol. 2, Berlin: Springer, 2010.
Luehr M, von Aspern K, Etz CD Limitations of Direct Regional Spinal Cord Monitoring Using Near-Infrared Spectroscopy: Indirect Paraspinal Collateral Network Surveillance Is the Answer!. Ann Thorac Surg. 2016 Mar;101(3):1238-9.
T. Neumuth, P. Jannin, J. Schlomberg, J. Meixensberger, P. Wiedemann, und O. Burgert, „Analysis of surgical intervention populations using generic surgical process models“, Int J Comput Assist Radiol Surg, Bd. 6, Nr. 1, S. 59–71, Jan. 2011. (a)
T. Neumuth, P. Jannin, G. Strauss, J. Meixensberger, und O. Burgert, „Validation of Knowledge Acquisition for Surgical Process Models“, J Am Med Inform Assoc, Bd. 16, Nr. 1, S. 72–80, 2009.
D. Neumuth, F. Loebe, H. Herre, und T. Neumuth, „Modeling surgical processes: a four-level translational approach“, Artif Intell Med, Bd. 51, Nr. 3, S. 147–161, März 2011. (b)
T. Neumuth, R. Wiedemann, C. Foja, P. Meier, J. Schlomberg, D. Neumuth, und P. Wiedemann, „Identification of surgeon-individual treatment profiles to support the provision of an optimum treatment service for cataract patients“, J Ocul Biol Dis Infor, Bd. 3, Nr. 2, S. 73–83, Apr. 2011. (c)
Pouw MH, A J F Hosman, J J van Middendorp, M M Verbeek, P E Vos and H van de Meent. Spinal Cord (2009) 47, 519–525; doi:10.1038/sc.2008.176; published online 20 January 2009. Biomarkers in spinal cord injury.
A.. Uciteli, S. Gross, H. Herre, “An Ontologically founded Architecture for Information Systems in Clinical and Epidemiological Research. “ BMC Biomedical Semantics, 2011.