AI Automated Thrombosis Detection

Revolutionising the Rapid Response to Thromboembolisms

Interaction of the blood tissue with a large areas of the artificial surface of the ECMO circuit contributes to the non-physiological blood flow that can increase shear stress and platelet activation [1]. Endothelial disruption at the site of catheter insertion combined with hypercoagulability and chances of venous stasis due to patient immobility, fulfils the Virchow's Triad. The resultant imbalance in homeostasis upregulates thrombosis characterised by elevated Factor Xa, fibrin and D-dimer.

The use of a centrifugal pump to control the blood flow can result in the displacement of the thrombus and increase incidences of venous thromboembolisms.

Studies from the Extracorporeal Life Support Organisation (ESLO) report thrombosis in 54.9% ECMO cases with a 10-16% chance of coagulation in the oxygenator module itself. [3]

Current manual methods of monitoring coagulatory biomarker levels is inefficient and can lead to human error, which can be fatal especially with ECMO being a last-resort therapy for patients.

Fig. 1. Contribution of Device and Patient Restrictions to the Fulfillment of Virchow's Triad [2]

Existing Safety Measures

To prevent thromboembolisms there are measure currently in place like the use of unfractionated heparin, which is considered the gold-standard anticogulant drug. This compound binds to the enzyme inhibitor of antithrombin III (ATIII), activatiing ATIII and inactivating coagulatory thrombin.

Periodic activated clotting time (ACT) is the point of care test that helps healthcare provider monitor the anticoagulatory effect of the UFH. The usual rate of infusion is infusion 20–50 units/kg/hour to attain the target ACT range of 180–220 seconds. Further, chromogenic tests for anti-Xa - the component inhibiting coagulatory factor Xa - have been incorporated into ECMO protocols with a desired range of 0.3–0.7 IU/mL of heparin for induction of anti-Xa activity.[3]

Additionally, pressure sensors are often implanted to measure the difference in arterial pressure (200 - 250 mm Hg) and venous (100 mm Hg) pressure. Any significant rise from normal values in the transmembrane pressure difference (> +20 mm Hg) raises an alert which stops the ECMO provided it is in 'intervention mode'. Normally, this testing is only indicative of some system malfunction and does not confirm the presence of a thrombus, requiring the doctor's interpretation and subsequent intervention. [4]

Depsite the development of state-of-the-art tubing systems with endothelium mimicking biocompatable surfaces with anti-thrombotic properties, the use of anticoagulants and immunosuppressants results in bleeding in up to 20–33% of ECMO patients, with increased incidence seen in venoarterial (V-A) ECMO. Statistically, bleeding events were associated with higher mortality (46.1%) than even thrombotic events (36.1%). [3]

Thus, the mechanisms in place to circumvent the incidence of thromboembolisms pose greater risks to the patient.

Fig. 2. Unusual Thrombus as Viewed in an ECMO System [4]

Non-invasive Duplex Ultrasonography

Fig. 3. Non-invasive Ultrasound Thrombotic Detection Set-up [5]

Leveraging the fact that ultrasonic velocity is higher in a thrombus than uncoagulated blood, recent studies have developed a non-invasive real-time ultrasonic test to detect an early floating thrombus. The combined use of a 5-MHz ultrasonic sensor, ultrasonic pulser-receiver, and Digital Storage Oscilloscope (DSO), enables the detection of changes in ultrasound flow rate, which can be specifically attributed to the thrombus. However, this solution is limited by the size of the thrombus, as it would not be able to detect pre-thrombotic events. [5]

Fig. 4. Left-side shift of waveform due to thrombus

Incorporating AI into Wearable Vein Checkers

With the advent of Artificial Intelligence and rise of automation, we can leverage AI prognostication for application to such complex therapies. Recent developments in Deep Vein Thrombosis (DVT) detection can be put to use in the ECMO system with suitable modifications. The incorporation of 3 testing mechanisms - ultrasound technology, impedance plethysmography (measuring changes in electrical resistance to estimate blood volume) and light reflection rheography (measuring changes in blood volume using infrared radiation)- into a small wearable device would improve the detection efficacy. Employing AI in the prediction of thrombotic events based on exisiting patient data is an avenue to explore for the mitigation of adverse thromboembolisms. This can also be integrated into a wearable device that can be prescribed even after weaning from the ECMO system, extending patient survival outcomes. [6]

Fig. 5. Impedance Plethysmography Illustration [8]

Fig. 6. Light Reflection Rheography Mechanism [9]

References

[1] Rajsic, S., Breitkopf, R., Jadzic, D., Popovic Krneta, M., Tauber, H., & Treml, B. (2022). Anticoagulation Strategies during Extracorporeal Membrane Oxygenation: A Narrative Review. Journal of clinical medicine, 11(17), 5147. https://doi.org/10.3390/jcm11175147 [2] Perepu, U.S. (2022). Antithrombotic Therapy for ECMO. In: Schmidt, G.A. (eds) Extracorporeal Membrane Oxygenation for Adults. Respiratory Medicine. Humana, Cham. https://doi.org/10.1007/978-3-031-05299-6_8 [3] Pacheco-Reyes, A. F., Plata-Menchaca, E. P., Mera, A., Bonilla, C., Sosa, M., Torrella, P., Palmada, C., Domènech, J., Rojas, S., Francés, S., Gallart, E., Girón, P., Nuvials, X., Ferrer, R., & Riera, J. (2024). Coagulation management in patients requiring extracorporeal membrane oxygenation support: A comprehensive narrative review. Annals of Blood, 9(0), Article 0. https://doi.org/10.21037/aob-23-28[4] Mossadegh, C. (2017). Monitoring the ECMO. In: Mossadegh, C., Combes, A. (eds) Nursing Care and ECMO. Springer, Cham. https://doi.org/10.1007/978-3-319-20101-6_5 [5] Rim, G., Hyun, K., Cho, D.G. et al. Early thrombus detection in the extracorporeal membrane oxygenation circuit by noninvasive real-time ultrasonic sensors. Sci Rep 14, 10438 (2024). https://doi.org/10.1038/s41598-024-59873-z [6] Seo, J.W., Park, S., Kim, Y.J. et al. Artificial intelligence-based iliofemoral deep venous thrombosis detection using a clinical approach. Sci Rep 13, 967 (2023). https://doi.org/10.1038/s41598-022-25849-0 [7] Stephens, A. F., Šeman, M., Diehl, A., Pilcher, D., Barbaro, R. P., Brodie, D., Pellegrino, V., Kaye, D. M., Gregory, S. D., Hodgson, C., & Extracorporeal Life Support Organization Member Centres (2023). ECMO PAL: using deep neural networks for survival prediction in venoarterial extracorporeal membrane oxygenation. Intensive care medicine, 49(9), 1090–1099. https://doi.org/10.1007/s00134-023-07157-x [8] Menon, J., Hamilton, G. (2007). impedance pleythography. Emergency Medicine Clinics of North America. Surgery (Oxford) [9] Arora, Subodh et al.(1993) Light reflection rheography: A simple noninvasive screening test for deep vein thrombosis. Journal of Vascular Surgery, Volume 18, Issue 5, 767 - 772

Shekdar Ira Nishith

Page updated

Google Sites

Report abuse