Arsalan Nisar, M.H.S.

Background: Acute ischemic stroke (AIS) comprise the overwhelming majority of hospitalized stroke cases, with more than 87% of all cases being attributed to AIS. Current diagnostic efforts consist mainly of non-contrast CT imaging and a comprehensive neurologic examination. However, due to the large number of stroke cases and limited physician support, the heavy reliance on imaging modalities increases the burden on physicians and creates numerous challenges in diagnosing such time sensitive ailments such as AIS. Thus, there is a need to increase current diagnostic capabilities through a supplementation with non-imaging based biomarkers and machine learning algorithms which may provide valuable information on the current state of a patient’s prognosis and the current severity of the condition. The author performed a literature review to analyze which biomarkers have been shown to be efficacious and what methods of machine learning and artificial intelligence have been implemented to reflect a correlation between levels of biomarkers and a patient’s health status as it pertains to AIS.

Objective: Discuss the current state of non-imaging based biomarkers and AI as a possible means to supplement current stroke diagnostic methods.

Search Methods: A literature review was conducted of primary research and mechanistic research articles through PubMed using search terms such as “Acute ischemic stroke”, “Biomarkers”, and “Diagnosis”. Peer review was undertaken to ascertain the validity of the findings and relevance to the topic.

Results: The results suggested that biomarkers, either alone or in combination with several others, can indicate a diagnosis of AIS and corresponding levels of severity. Some approaches used singular biomarker levels, such as Gelsolin, to directly indicate AIS diagnosis (2). Other approaches utilized single metabolites, such as Phenylacetylglutamine (PAG), and estimated their contribution to variables such as modified Rankin Scale – a measure of disability severity in AIS (5). Finally, other approaches used multiple biomarker level changes and ratios to have more robust diagnostic potential (4).

In addition to biomarkers, another approach that is promising is to incorporate AI into the diagnostic process.  Many previous approaches have limited clinical significance as there was no use of clinically relevant diagnostic models (1). Artificial Intelligence (AI) has shown increasing capabilities to fill this gap and provide significant clinical diagnostic potential particularly in the realm of AIS. For example, a recent study illustrated the efficacy of Artificial Neural Networks (ANN) and Machine Learning Models (MLM) in predicting AIS and disease progression using miRNAs as a biological biomarker (5).

Conclusion: There are several instances by which AIS status can be assessed by non-imaging based biomarkers. Methods include single biomarkers, multiple biomarkers, and changes in biomarker levels. AI, such as ANNs and MLMs, have shown promise to assess patients with prospective data from levels of biomarkers and diagnose AIS in patients with 94.8% and 95.8% sensitivity. In conclusion, AI has increased capacity to serve as a method to develop diagnostic models for AIS and further investigation should be done with biomarkers that have significant clinical correlation potential.

 Works Cited:

1.  Fasen, B. A. C. M., Berendsen, R. C. M., & Kwee, R. M. (2021). Artificial intelligence software for diagnosing intracranial arterial occlusion in patients with acute ischemic stroke. American Journal of Neuroradiology, 42(4), 597-603. doi: 10.3174/ajnr.A700
2. García-Berrocoso T, Penalba A, Boada C, Giralt D, Cuadrado E, Colomé N, Dayon L, Canals F, Sanchez JC, Rosell A, Montaner J. From brain to blood: New biomarkers for ischemic stroke prognosis. Annals of Neurology. 2012 Aug;72(2):363-75. doi: 10.1002/ana.23647. PMID: 22926855.
3. Ovbiagele B, Nguyen-Huynh MN. Stroke epidemiology: advancing our understanding of disease mechanism and therapy. Neurotherapeutics. 2011 Jul;8(3):319-29. doi: 10.1007/s13311-011-0053-1. PMID: 21691873; PMCID: PMC3250269.
4. Smith WS, Sung G, Saver J, Budzik R, Duckwiler G, Liebeskind DS, Lutsep HL, Rymer MM, Higashida RT, Starkman S, Gobin YP; Multi MERCI Investigators; Frei D, Grobelny T, Hellinger F, Huddle D, Kidwell C, Koroshetz W, Marks M, Nesbit G, Silverman IE. Mechanical thrombectomy for acute ischemic stroke: final results of the Multi MERCI trial. Stroke. 2008 Apr;39(4):1205-12. doi: 10.1161/STROKEAHA.107.497115. Epub 2008 Feb 28. PMID: 18309168.
5. Sun, R., Li, Y., Cai, M., Cao, Y., & Piao, X. (2020). Discovery of a New Biomarker Pattern for Differential Diagnosis of Acute Ischemic Stroke Using Targeted Metabolomics. Journal of Stroke and Cerebrovascular Diseases, 29(11), 105173. Doi: 10.1016/j.jstrokecerebrovasdis.2020.105173.
6. Yu F, Li X, Feng X, Wei M, Luo Y, Zhao T, Xiao B and Xia J (2022) Phenylacetylglutamine, a Novel Biomarker in Acute Ischemic Stroke. Front. Cardiovasc. Med. 8:798765. doi: 10.3389/fcvm.2021.798765
7. Xinyi Zhao, Xingmei Chen, Xulong Wu, Lulu Zhu, Jianxiong Long, Li Su, Lian Gu (2021). Machine Learning Analysis of MicroRNA Expression Data Reveals Novel Diagnostic Biomarker for Ischemic Stroke. Journal of Stroke and Cerebrovascular Diseases, Volume 30, Issue 8, 2021,105825. ISSN 1052-3057. https://doi.org/10.1016/j.jstrokecerebrovasdis.2021.105825.