Periodic Reporting for period 3 - INSIST (IN-Silico trials for treatment of acute Ischemic STroke)
Berichtszeitraum: 2020-11-01 bis 2022-04-30
Stroke is a major health problem in the Western world; 1.3 million Europeans have a first stroke each year, 30% die and 30% are left permanently disabled. One in 5 patients require long term institutional care, costing ~€38 billion per year in Europe. An acute ischemic stroke occurs when an artery supplying blood to the brain is blocked by a blood clot (thrombus), causing rapid loss of brain function. Treatment is limited to thrombolysis by clot-dissolving enzymes and thrombectomy by mechanical devices. As still 2 out of 3 patients become functionally dependent, further improvement of treatment is needed. The required clinical trials are extremely expensive.
In silico models of patient characteristics, pathophysiology and (effects of) treatment may be used to set up in silico clinical trials (“virtual” patients are “virtually” treated), potentially reducing, refining and partially replacing actual clinical trials. This will allow testing efficacy of new treatments, suitability for specific patient(s) populations, and provide tools for in silico proof of concept trial design modelling, resulting in faster and safer marketing of new drugs and devices.
There are 5 objectives (O1-5) outlined in the section below.
In silico models of patient characteristics, pathophysiology and (effects of) treatment may be used to set up in silico clinical trials (“virtual” patients are “virtually” treated), potentially reducing, refining and partially replacing actual clinical trials. This will allow testing efficacy of new treatments, suitability for specific patient(s) populations, and provide tools for in silico proof of concept trial design modelling, resulting in faster and safer marketing of new drugs and devices.
There are 5 objectives (O1-5) outlined in the section below.
O1: Virtual populations
The final public version of the virtual population is launched and publicly available as online dashboard (https://mdmtest.shinyapps.io/INSIST-VP/) enabling users to visualize and download virtual stroke patient cohorts according to their characteristics of interest (i.e. inclusion and exclusion criteria of a study population). A final statistical model (module III) that relates image outcome to clinical outcome was also developed and the performance of this final model was assessed with internal and external validation. Module III includes two statistical models, which can accurately predict functional outcome at 3 months and NIHSS 24-48 hours after EVT for AIS.
O2: thrombosis & thrombolysis
The goal of WP3 was to develop numerical models for thrombosis and thrombolysis, driven by the objectives of simplicity and computational efficiency, while providing a quantitative description of the phenomena and a predictive approach. We developed a 3D Lattice-Boltzmann solver with hemodynamic capabilities and simplified lysis mechanisms. The main effort during this last period was the validation of the numerical models and their improvements. In terms of thrombolysis, an innovative experiment was developed to study the lysis process in a flow situation which revealed several unknown features that were included in our 3D thrombolysis model. They gave rise to an analytical surrogate model giving the speed of lysis as a function of the biophysical conditions.
O3: thrombectomy
A new constitutive law was developed for non-linear inelastic behaviour of blood clots based on deformation induced bond formation and dissociation of the fibrin network. Lap shear fracture tests and cohesive zone fracture modelling was performed to simulate mixed-mode crack propagation in blood clots. 100 patient-specific vascular models were reconstructed from computer tomography images of stroke patients. For each case in-silico thrombectomy simulations were performed to predict the outcomes of two stent designs (TREVO and EmboTrap). A high success rate was predicted for both devices (over 80%). A combined in silico mechanical thrombectomy (MT) with stent-retriever and aspiration catheters, both balloon guided catheter and double aspiration catheter, is developed and validated with in vitro experiments. Three different MT procedures were performed in the same vessel-like 3D-printed silicone phantom. The 3D phantom was designed in order to replicate the physiological dimension of the cerebral arteries. The model truthfully replicated the experimentally measured modes of thrombus fragmentation during thrombectomy retrieval.
O4: perfusion defects & tissue damage
In WP5, the focus of the final period of the project has been to provide detailed validation of the models of blood flow and perfusion, using both novel animal data and large clinical datasets. This has resulted in the most detailed and highly validated model of its type in the world for the effects of microthrombi on perfusion and tissue viability. In particular, the multi-scale approach has enabled us to link together both newly acquired animal data and clinical data within the same modelling framework.
O5: stroke trial platform & in silico clinical trial
The above work culminates here. We created an INSIST ISCT Computing environment. with a python command line interface, and parallel capabilities to scale up the cohort sizes that can be simulated in the trials. All final delivered versions of the WP2–5 models were integrated and EasyVVUQ has been integrated with the INSIST framework to enable easy implementation of VVUQ methodologies. EasyVVUQ has been integrated with the INSIST framework to enable easy implementation of VVUQ methodologies. A validation study has been performed using the MR CLEAN Trial clinical data but found the trial is currently only credible for use in testing thrombectomy surgical outcome on populations with M1 occlusions. In addition to the validation trial, two exploratory trials were run: comparing thrombus composition and thrombectomy device. Given the results of the validation, these two trials focussed on the outcome of the thrombectomy procedure.
The final public version of the virtual population is launched and publicly available as online dashboard (https://mdmtest.shinyapps.io/INSIST-VP/) enabling users to visualize and download virtual stroke patient cohorts according to their characteristics of interest (i.e. inclusion and exclusion criteria of a study population). A final statistical model (module III) that relates image outcome to clinical outcome was also developed and the performance of this final model was assessed with internal and external validation. Module III includes two statistical models, which can accurately predict functional outcome at 3 months and NIHSS 24-48 hours after EVT for AIS.
O2: thrombosis & thrombolysis
The goal of WP3 was to develop numerical models for thrombosis and thrombolysis, driven by the objectives of simplicity and computational efficiency, while providing a quantitative description of the phenomena and a predictive approach. We developed a 3D Lattice-Boltzmann solver with hemodynamic capabilities and simplified lysis mechanisms. The main effort during this last period was the validation of the numerical models and their improvements. In terms of thrombolysis, an innovative experiment was developed to study the lysis process in a flow situation which revealed several unknown features that were included in our 3D thrombolysis model. They gave rise to an analytical surrogate model giving the speed of lysis as a function of the biophysical conditions.
O3: thrombectomy
A new constitutive law was developed for non-linear inelastic behaviour of blood clots based on deformation induced bond formation and dissociation of the fibrin network. Lap shear fracture tests and cohesive zone fracture modelling was performed to simulate mixed-mode crack propagation in blood clots. 100 patient-specific vascular models were reconstructed from computer tomography images of stroke patients. For each case in-silico thrombectomy simulations were performed to predict the outcomes of two stent designs (TREVO and EmboTrap). A high success rate was predicted for both devices (over 80%). A combined in silico mechanical thrombectomy (MT) with stent-retriever and aspiration catheters, both balloon guided catheter and double aspiration catheter, is developed and validated with in vitro experiments. Three different MT procedures were performed in the same vessel-like 3D-printed silicone phantom. The 3D phantom was designed in order to replicate the physiological dimension of the cerebral arteries. The model truthfully replicated the experimentally measured modes of thrombus fragmentation during thrombectomy retrieval.
O4: perfusion defects & tissue damage
In WP5, the focus of the final period of the project has been to provide detailed validation of the models of blood flow and perfusion, using both novel animal data and large clinical datasets. This has resulted in the most detailed and highly validated model of its type in the world for the effects of microthrombi on perfusion and tissue viability. In particular, the multi-scale approach has enabled us to link together both newly acquired animal data and clinical data within the same modelling framework.
O5: stroke trial platform & in silico clinical trial
The above work culminates here. We created an INSIST ISCT Computing environment. with a python command line interface, and parallel capabilities to scale up the cohort sizes that can be simulated in the trials. All final delivered versions of the WP2–5 models were integrated and EasyVVUQ has been integrated with the INSIST framework to enable easy implementation of VVUQ methodologies. EasyVVUQ has been integrated with the INSIST framework to enable easy implementation of VVUQ methodologies. A validation study has been performed using the MR CLEAN Trial clinical data but found the trial is currently only credible for use in testing thrombectomy surgical outcome on populations with M1 occlusions. In addition to the validation trial, two exploratory trials were run: comparing thrombus composition and thrombectomy device. Given the results of the validation, these two trials focussed on the outcome of the thrombectomy procedure.
The population model and the virtual populations of stroke patients with both clinical and imaging data of acute ischemic stroke is a valuable data source for research. We created a public version of our virtual stroke patient model, available through a web application (https://mdmtest.shinyapps.io/INSIST-VP/).
Virtual populations have been used for ASCLEPIOS, a security-oriented Horizon program and for a hackathon for which also virtual but realistic data of patients with acute ischemic stroke were required. We are currently exploring multiple sources of funding to continue providing the service of the in-silico modelling of stroke treatment and disease.
The exploitation of the INSIST models takes quite some time. Currently there is quite some interest from companies like Certara, Ansys, Philips, and LifeTec. However, for these companies, more work is required to realize adoption of the INSIST technology. However, the Technology Transfer Office from the AMC has initiated contact with the other partners to discuss the terms of use of our technology.
We have explored additional funding schemes for future extension of the INSIST population models and virtual populations. In doing so, we have set up a collaboration with Syntho (http://www.syntho.com ), a company providing a service of generating synthetic data from existing databases. We have explored funding opportunities and found interesting possibilities within the NLAIC.
Finally, to popularize in silico trials for development of new medical devices, drugs, or treatment procedures, we presented the modelling framework des-ist (Discrete Event Simulation framework for In Silico Trials1). This framework supports discrete event-based simulations. An example in silico trial was highlighted studying treatment of acute ischemic stroke, as considered in INSIST. The GitHub project for the desist software (i.e. the event based framework for running in silico trials) is freely available at: https://github.com/UvaCsl/desist
Virtual populations have been used for ASCLEPIOS, a security-oriented Horizon program and for a hackathon for which also virtual but realistic data of patients with acute ischemic stroke were required. We are currently exploring multiple sources of funding to continue providing the service of the in-silico modelling of stroke treatment and disease.
The exploitation of the INSIST models takes quite some time. Currently there is quite some interest from companies like Certara, Ansys, Philips, and LifeTec. However, for these companies, more work is required to realize adoption of the INSIST technology. However, the Technology Transfer Office from the AMC has initiated contact with the other partners to discuss the terms of use of our technology.
We have explored additional funding schemes for future extension of the INSIST population models and virtual populations. In doing so, we have set up a collaboration with Syntho (http://www.syntho.com ), a company providing a service of generating synthetic data from existing databases. We have explored funding opportunities and found interesting possibilities within the NLAIC.
Finally, to popularize in silico trials for development of new medical devices, drugs, or treatment procedures, we presented the modelling framework des-ist (Discrete Event Simulation framework for In Silico Trials1). This framework supports discrete event-based simulations. An example in silico trial was highlighted studying treatment of acute ischemic stroke, as considered in INSIST. The GitHub project for the desist software (i.e. the event based framework for running in silico trials) is freely available at: https://github.com/UvaCsl/desist