Automated Diagnosis of Stroke in Computed Tomography With the Use of Artificial Intelligence

Overview

The use of the systems of machine learning for the quantification, location and diagnosis of ischemic stroke in non-contrasted head computed tomography, is a method with high efficacy, accessible and susceptible for standardization, for the assistance in the clinical decision making in the absence of specialized health personnel for the attention of this disease.

Full Title of Study: “Assisted Method for the Quantification, Localization and Automated Diagnosis of Stroke in Computed Tomography With the Use of Artificial Intelligence”

Study Type

  • Study Type: Observational [Patient Registry]
  • Study Design
    • Time Perspective: Prospective
  • Study Primary Completion Date: June 1, 2025

Detailed Description

PROBLEM STATEMENT Stroke is the fifth cause of death in the United States and the number one cause of long term disability. Annually around 800,000 persons are diagnosed with stroke in this country, causing the death of around 130,000 persons, each 4 minutes one person die due to this disease in the U.S.A. In Mexico there is very few statistical information published about stroke, according to the Mexican clinical practice guidelines "CENETEC" this disease was the third cause of death, with the 5.6% percent of the main death causes, with an estimated rate of 25.6/100,000 habitants and more than 25,000 deaths per year according to data of the health department. The monetary cost of the direct attention and the indirect expenses is enormous, with an approximate expenditure of 34 billion dollars per year. Around 20% of the survivors require special care during the first 3 months after a stroke event and around 30% remain with a severe disability. In Mexico there is an alarming increase of stroke cases that correlates with the demographic transition to an older population, and also to the increase of risk factors for vascular atherothrombotic disease, as the hypertension, diabetes, obesity and dyslipidemia. Local statistics show that from the different type of stroke the distribution in Mexico is: 72.94% as ischemic stroke, 20.17% hemorrhagic and 6.8% with subarachnoid hemorrhage, this numbers are similar to the statistic reported in the literature of other countries. According to the first multicentric registry in Mexico "The Premier Study" the distribution of the etiology of the ischemic stroke was: large vessel atherosclerosis 8%, cardiac embolism 20%, lacunar stroke 20% , miscellaneous 5% and undetermined etiology 41%. In Mexico only 0.6% of the patients with stroke are treated with I.V. thrombolysis, and of these patients, 33% arrived in a window <3 hrs. after the onset of the stroke. JUSTIFICATION The medical care and diagnosis of the patients with stroke is considered a medical emergency, studies have reported the denominated "Time is brain" due to the fact, that patients with a typical large vessel ischemic stroke, each minute 1.9 million neurons, 14 billion of synapsis and around 12 km. of myelinated fibers are lost, this remark the need of an early diagnosis and treatment. According to "The Premier Study" one of the most concerning outcomes in the medical care of stroke in Mexico is that only a fifth of the patients with stroke arrive to the hospital within the window for I.V. thrombolysis, and of them < 1% receive treatment with rTPA. This is due to factors as the lack of specialized medical teams for an opportune and precise diagnosis, a limited number of staff with expertise in stroke and a lack of technological resources for diagnosis. Fort he particular case of the stroke, there are several characteristics that propitiate the use of artificial intelligence (A.I.) systems, some of them are the great quantity of data and multidisciplinary approaches used for the medical decision making, particularly the use of imaging studies, due to the fact that this are key factors in the stroke management. Other factors that have contributed for the use of A.I. in stroke are the lack of experts for the imaging diagnosis, this problem is clearly exacerbated in developing countries. Only a few hospitals have access to specialized neuroradiologists or neurologist to provide an opportune and optimal diagnosis, this generates a delay or even an inaccurate diagnosis. There is even a variable concordance ranging up to 39% between specialized physicians for the diagnosis of stroke in non-contrasted computed tomography in the medial cerebral artery territory, this demonstrate the great inter-reader variability, furthermore there is sparse definition of the characteristics of early infarct, demonstrating the need for a standardized method of stroke imaging. Several studies have demonstrated the efficacy of the A.I. systems in the stroke diagnosis, most of them show that the automatized diagnosis with the A.I. is in the majority of the times the same or even better than the manual segmentation with human diagnosis by stroke experts. There is even a stroke evaluation system already in the market, based on A.I. algorithms, this has already demonstrated its utility in the ENCHANTED trial (Enhanced Control of Hypertension and Thrombolysis Stroke Study), "The Electronic-Alberta Stroke Program Early Computed Tomography Score Software. BACKGROUNDS The A.I. has been defined as the study of the computations that make possible to perceive, think and act. A.I. systems have focused to emulate the cognitive functions of the humans, due to the rapidly advances in the computational sciences the A.I. is being used in multiple fields, especially in big companies of information technology as google. The precision medicine has been defined as the personalized diagnosis and treatment of an individual patient, this is an emerging concept in the contemporary medicine. The great number of information generated in biology, immunology, genomics and the advances in the imaging technology has created the path to this personalized medicine. These advances have generated a transition in the modern healthcare models, propelled by the great quantity of data and the advance of the analytic techniques. Guided by relevant clinical questions, the A.I. systems can reveal valuable information hided within massive amounts of data (big data), due to its analytic qualities. This can be helpful in the clinical decision making. Currently there are companies as I.B.M. and it's A.I. system Watson for Oncology (WFO), developing Clinical Decision-Support Systems (CDSSs) in collaboration with hospitals as the Memorial Sloan Kettering Cancer Center, this system pretends to help oncology specialist with the current challenges of the great amount of data in fields as genetics, pharmacology and treatment guidelines, analyzing big data bases for a more personalized treatment of the patient. WFO was tested with real cases discussed in the multidisciplinary boards at the Memorial Sloan Kettering Cancer Center obtaining a great level of concordance in their series, and currently is already available in the market. HYPOTHESIS The use of the systems of machine learning for the quantification, location and diagnosis of ischemic stroke in non-contrasted head computed tomography, is a method with high efficacy, accessible and susceptible for standardization, for the assistance in the clinical decision making in the absence of specialized health personnel for the attention of this disease. OBJECTIVES General Objective • Design an A.I. algorithm for the quantification, location and diagnosis of ischemic stroke in non-contrasted head computed tomography, base in a hybrid model with the use of machine learning and statistic modeling, evaluated its efficacy and create a software for the assistance of medical decision making. Specific objectives – Standardize the data collection and the creation of data banks of non-contrasted head computed tomography with ischemic stroke diagnosis for the inclusion into the A.I. – Create non-contrasted head computed tomography data bases with representative samples of the different types of ischemic stroke, considering the multiple anatomical distributions, sizes and evolution phases. – Standardize an image segmentation method, location of the lesion, diagnosis and ASPECT (Alberta Stroke Programme Early CT Score) scoring. – Generate an efficient software for the assistance of clinical decision making to non-specialized medical staff in the diagnosis of this disease, in order to expedite the medical care of this population of patients in places, where a lack of specialized teams and technologic resources exists, for the diagnosis of this disease. Technical Objectives – Standardize the collection of the anonymized non-contrasted head computed tomography DICOM files with their respective system specifications (headers) for the creation of the "Ground Truth" of the machine learning system. – Create an efficient algorithm and evaluate its accuracy in the stroke diagnosis through several pre-clinical tests with expert specialized physicians in cerebrovascular disease. – Design an interactive software and easy to use for the medical staff, for fast a clinical decision making and with support on evidence-based medicine for patients with suspicion of ischemic stroke. GOALS – Create a software for the quantification, location and automatized diagnosis of stroke , that is efficient and accurate for the clinical decision making in the absence of specialized medical staff and technological resources for the medical care of patients with stroke. – Obtain the register of a patent of the machine learning based software. – Generate the publication of multiple articles in international journals. METHODOLOGY Type of study Observational prospective study, descriptive, without patient intervention, without blinding for the analysis of the non-contrasted head computed tomography and its clinical correlation. In the present study the confidentiality of the patient will be maintained, as the investigators won't need the identification data from the patients. In the present study the investigators pretend to create a hybrid A.I. model using a combined method of statistic modeling for the first prototype of the algorithm and then continue with the creation of the machine learning algorithm, combining both algorithms for a better accuracy and efficacy. For the first algorithm with statistic modeling the investigators pretend to obtain a representative sample with the different types of ischemic stroke, until the investigators collect 50 images, clinical information and its respective diagnosis and ASPECT score. Continuing with the prospective creation of a data base of computed tomography imaging, clinical correlation, and their respective diagnosis and ASPECT score for the feeding of the machine learning algorithm, that usually requires large quantities of data for its training. For both algorithms the investigators pretend to test them with independent evaluators, whose expertise will be on cerebrovascular disease, and the investigators will compare de results of the A.I. system with the diagnosis of the specialized evaluators, in order to calibrate the cut off points of the algorithm to obtain the maximal concordance . INFRASTRUCTURE For the developing of this protocol the investigators pretend to use the facilities of the Department of Neurosciences form the Health Sciences Center of the University of Guadalajara "CUCS". The collection of the images will be done in collaboration with the Department of radiology of the "Hospital Civil de Guadalajara, Fray Antonio Alcalde" and "Hospital Civil de Guadalajara, Juan I. Menchaca". The evaluation of the images will be performed in collaboration with the Neurology Department of the "Hospital Civil de Guadalajara, Fray Antonio Alcalde". The creation of the A.I. algorithm will be done in collaboration with the University Center of Engineering and Exact Sciences (CUCEI) from the University of Guadalajara, in the department of computational sciences from CUCEI.

Arms, Groups and Cohorts

  • Ischemic stroke patients with <4.5 hours of onset .
    • analysis of the computed tomography without contrast and detection of early changes, scoring with ASPECTS score and integration to the machine learning data base.

Clinical Trial Outcome Measures

Primary Measures

  • Alberta Stroke Program Early Computed Tomography Score
    • Time Frame: 1 day
    • The Alberta stroke programme early CT score (ASPECTS) 1 is a 10-point quantitative topographic CT scan score used in patients with middle cerebral artery stroke. It has also been adapted for the posterior circulation. Scoring system Segmental assessment of the MCA vascular territory is made and 1 point is deducted from the initial score of 10 for every region involved: caudate putamen internal capsule insular cortex M1: “anterior MCA cortex,” corresponding to frontal operculum M2: “MCA cortex lateral to insular ribbon” corresponding to anterior temporal lobe M3: “posterior MCA cortex” corresponding to posterior temporal lobe M4: “anterior MCA territory immediately superior to M1” M5: “lateral MCA territory immediately superior to M2” M6: “posterior MCA territory immediately superior to M3”

Participating in This Clinical Trial

Inclusion Criteria

  • Patients with ischemic stroke. – Patients with <4.5 hours of symptom onset. Exclusion Criteria:

  • Patients with hemorrhagic stroke diagnosis. – Patients with other cerebrovascular diseases.

Gender Eligibility: All

Minimum Age: 18 Years

Maximum Age: N/A

Are Healthy Volunteers Accepted: No

Investigator Details

  • Lead Sponsor
    • University of Guadalajara
  • Provider of Information About this Clinical Study
    • Principal Investigator: Ivan Segura Duran, Principal Investigator – University of Guadalajara
  • Overall Official(s)
    • Rodrigo Ramos Zúñiga, M.D. Ph.D., Principal Investigator, University of Guadalajara
    • Eduardo Gerardo Mendizábal Ruiz, Ph.D., Principal Investigator, University of Guadalajara
    • José Luis Ruiz Sandoval, M.D., Principal Investigator, University of Guadalajara
    • Ivan Segura Duran, M.D., Study Chair, University of Guadalajara

References

Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman M, de Ferranti S, Despres JP, Fullerton HJ, Howard VJ, Huffman MD, Judd SE, Kissela BM, Lackland DT, Lichtman JH, Lisabeth LD, Liu S, Mackey RH, Matchar DB, McGuire DK, Mohler ER 3rd, Moy CS, Muntner P, Mussolino ME, Nasir K, Neumar RW, Nichol G, Palaniappan L, Pandey DK, Reeves MJ, Rodriguez CJ, Sorlie PD, Stein J, Towfighi A, Turan TN, Virani SS, Willey JZ, Woo D, Yeh RW, Turner MB; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics–2015 update: a report from the American Heart Association. Circulation. 2015 Jan 27;131(4):e29-322. doi: 10.1161/CIR.0000000000000152. Epub 2014 Dec 17. No abstract available. Erratum In: Circulation. 2015 Jun 16;131(24):e535. Circulation. 2016 Feb 23;133(8):e417.

Benjamin EJ, Blaha MJ, Chiuve SE, Cushman M, Das SR, Deo R, de Ferranti SD, Floyd J, Fornage M, Gillespie C, Isasi CR, Jimenez MC, Jordan LC, Judd SE, Lackland D, Lichtman JH, Lisabeth L, Liu S, Longenecker CT, Mackey RH, Matsushita K, Mozaffarian D, Mussolino ME, Nasir K, Neumar RW, Palaniappan L, Pandey DK, Thiagarajan RR, Reeves MJ, Ritchey M, Rodriguez CJ, Roth GA, Rosamond WD, Sasson C, Towfighi A, Tsao CW, Turner MB, Virani SS, Voeks JH, Willey JZ, Wilkins JT, Wu JH, Alger HM, Wong SS, Muntner P; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart Disease and Stroke Statistics-2017 Update: A Report From the American Heart Association. Circulation. 2017 Mar 7;135(10):e146-e603. doi: 10.1161/CIR.0000000000000485. Epub 2017 Jan 25. No abstract available. Erratum In: Circulation. 2017 Mar 7;135(10 ):e646. Circulation. 2017 Sep 5;136(10 ):e196.

Saver JL. Time is brain–quantified. Stroke. 2006 Jan;37(1):263-6. doi: 10.1161/01.STR.0000196957.55928.ab. Epub 2005 Dec 8.

Grotta JC, Chiu D, Lu M, Patel S, Levine SR, Tilley BC, Brott TG, Haley EC Jr, Lyden PD, Kothari R, Frankel M, Lewandowski CA, Libman R, Kwiatkowski T, Broderick JP, Marler JR, Corrigan J, Huff S, Mitsias P, Talati S, Tanne D. Agreement and variability in the interpretation of early CT changes in stroke patients qualifying for intravenous rtPA therapy. Stroke. 1999 Aug;30(8):1528-33. doi: 10.1161/01.str.30.8.1528.

Mullins ME, Lev MH, Schellingerhout D, Koroshetz WJ, Gonzalez RG. Influence of availability of clinical history on detection of early stroke using unenhanced CT and diffusion-weighted MR imaging. AJR Am J Roentgenol. 2002 Jul;179(1):223-8. doi: 10.2214/ajr.179.1.1790223.

Nagel S, Wang X, Carcel C, Robinson T, Lindley RI, Chalmers J, Anderson CS; ENCHANTED Investigators. Clinical Utility of Electronic Alberta Stroke Program Early Computed Tomography Score Software in the ENCHANTED Trial Database. Stroke. 2018 Jun;49(6):1407-1411. doi: 10.1161/STROKEAHA.117.019863. Epub 2018 May 18.

Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, Wang Y, Dong Q, Shen H, Wang Y. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. 2017 Jun 21;2(4):230-243. doi: 10.1136/svn-2017-000101. eCollection 2017 Dec.

Somashekhar SP, Sepulveda MJ, Puglielli S, Norden AD, Shortliffe EH, Rohit Kumar C, Rauthan A, Arun Kumar N, Patil P, Rhee K, Ramya Y. Watson for Oncology and breast cancer treatment recommendations: agreement with an expert multidisciplinary tumor board. Ann Oncol. 2018 Feb 1;29(2):418-423. doi: 10.1093/annonc/mdx781.

Citations Reporting on Results

Cantu-Brito C, Ruiz-Sandoval JL, Murillo-Bonilla LM, Chiquete E, Leon-Jimenez C, Arauz A, Villarreal-Careaga J, Barinagarrementeria F, Ramos-Moreno A; PREMIER Investigators. The first Mexican multicenter register on ischaemic stroke (the PREMIER study): demographics, risk factors and outcome. Int J Stroke. 2011 Feb;6(1):93-4. doi: 10.1111/j.1747-4949.2010.00549.x. No abstract available.

Cantu-Brito C, Ruiz-Sandoval JL, Murillo-Bonilla LM, Chiquete E, Leon-Jimenez C, Arauz A, Villarreal-Careaga J, Rangel-Guerra R, Ramos-Moreno A, Barinagarrementeria F; PREMIER Investigators. Acute care and one-year outcome of Mexican patients with first-ever acute ischemic stroke: the PREMIER study. Rev Neurol. 2010 Dec 1;51(11):641-9. English, Spanish.

Lee EJ, Kim YH, Kim N, Kang DW. Deep into the Brain: Artificial Intelligence in Stroke Imaging. J Stroke. 2017 Sep;19(3):277-285. doi: 10.5853/jos.2017.02054. Epub 2017 Sep 29.

Clinical trials entries are delivered from the US National Institutes of Health and are not reviewed separately by this site. Please see the identifier information above for retrieving further details from the government database.

At TrialBulletin.com, we keep tabs on over 200,000 clinical trials in the US and abroad, using medical data supplied directly by the US National Institutes of Health. Please see the About and Contact page for details.