Artificial Intelligence in Medicine and Cardiac Imaging: Harnessing Big Data and Advanced Computing to Provide Personalized Medical Diagnosis and Treatment

  • Published on
    22-Dec-2016

  • View
    223

  • Download
    0

Transcript

NUCLEAR CARDIOLOGY (V DILSIZIAN, SECTION EDITOR)Artificial Intelligence in Medicine and Cardiac Imaging:Harnessing Big Data and Advanced Computing to ProvidePersonalized Medical Diagnosis and TreatmentSteven E. Dilsizian & Eliot L. SiegelPublished online: 13 December 2013# Springer Science+Business Media New York (outside the USA) 2013Abstract Although advances in information technology inthe past decade have come in quantum leaps in nearly everyaspect of our lives, they seem to be coming at a slower pace inthe field of medicine. However, the implementation of elec-tronic health records (EHR) in hospitals is increasing rapidly,accelerated by the meaningful use initiatives associated withthe Center for Medicare & Medicaid Services EHR IncentivePrograms. The transition to electronic medical records andavailability of patient data has been associated with increasesin the volume and complexity of patient information, as wellas an increase in medical alerts, with resulting alert fatigueand increased expectations for rapid and accurate diagnosisand treatment. Unfortunately, these increased demands onhealth care providers create greater risk for diagnostic andtherapeutic errors. In the near future, artificial intelligence(AI)/machine learning will likely assist physicians with dif-ferential diagnosis of disease, treatment options suggestions,and recommendations, and, in the case of medical imaging,with cues in image interpretation. Mining and advanced anal-ysis of big data in health care provide the potential not onlyto perform in silico research but also to provide real timediagnostic and (potentially) therapeutic recommendationsbased on empirical data. On demand access to high-performance computing and large health care databases willsupport and sustain our ability to achieve personalized medi-cine. The IBM Jeopardy! Challenge, which pitted the best all-time human players against the Watson computer, capturedthe imagination of millions of people across the world anddemonstrated the potential to apply AI approaches to a widevariety of subject matter, including medicine. The combina-tion of AI, big data, and massively parallel computing offersthe potential to create a revolutionary way of practicingevidence-based, personalized medicine.Keywords Artificial intelligence . Big data . Personalizedmedicine . IBMsWatson . Electronic health records . Neuralnetworks . Cardiac imagingIntroductionAll medical knowledgeincluding the continuous addition ofnew and important scientific informationcannot be proc-essed and stored by a single human brain. Physicians learnthousands of different diseases in medical school and areexpected to remember and apply a substantial subset of thesein daily practice. But it is impossible for an individual physi-cian to keep current on the broad spectrum of new data anddiscoveries and to reliably recall and utilize that information atall relevant time points. This is part of a major challenge inmedical imaging, where real-time errors are estimated toaverage between 3 % and 5 % and constitute nearly 75 % ofmedical malpractice claims [1]. Graber et al. estimated thatapproximately 75 % of diagnostic errors were related tocognitive factors [2]. Diagnostic errors outnumber othermedical errors by 2- to 4-fold and represent nearly 40 % oftotal ambulatory malpractice claims [3].These cognitive errors include anchoring bias (being stuckon an initial impression), framing bias or faulty context gen-eration (over-reliance on the specific way in which a questionis posed), availability bias (tendency to jump to a conclusionbased on a recent incident), satisfaction of search (not consid-ering other possibilities once a probable answer is found), andThis article is part of the Topical Collection on Nuclear CardiologyS. E. Dilsizian : E. L. SiegelUniversity of Maryland School of Medicine, Baltimore, MD, USAE. L. Siegel (*)Imaging Services, VA Maryland Health Care System, 10 NorthGreene Street, Mail Code 114, Baltimore, MD 21201, USAe-mail: esiegel@umaryland.eduCurr Cardiol Rep (2014) 16:441DOI 10.1007/s11886-013-0441-8premature closure (acceptance of an answer before it isverified) [1]. Cognitive errors, such as premature closure andfaulty context generation, have been implicated in 75 % ofpatient deaths in which physician/medical error was thoughtto play a role [4]. Machine learning and other artificialintelligence [AI] systems have the potential to be less sus-ceptible to these biases and, despite their limitations, can servein a complementary role to human decision makers.According to a study from researchers at Johns HopkinsUniversity, ~40,500 patients die in intensive care units eachyear in the United States as a result of diagnostic errors [5].System-related factors, such as poor processes, teamwork, andcommunication were involved in 65 % of these cases. Thesetypes of diagnostic problems contribute significantly to risinghealth care costs, with an estimated /300,000 per malpracticeclaim for misdiagnosis resulting from cognitive error orsystem-related factors [4]. It is anticipated that routine appli-cation of AI will decrease the risk of such errors. Such ad-vanced technology may ultimately replace a substantial per-centagealthough certainly not allof the work physiciansdo on a daily basis.Effective, evidence-based medical practice requires thatphysicians be familiar with the most recent guidelines andappropriate use criteria. Because of the exponentially growingamount of information in peer-review journals, textbooks,periodicals, consensus panels, and other sources, it is impos-sible for health care practitioners to keep up with more than asmall fraction of relevant literature. Adherence to guidelinesand evidence-based medicine may be made even more com-plex by the variability in standards of practice across differ-ent communities and states, a variability that complicates theconcept of a gold standard for diagnosis and treatment ofcertain illnesses. Advanced computer systems, such as IBMsWatson technology, could assist by providing the most up-to-date evidence-based information to inform proper patientcare decisions. This information could combine data from aspecific patients history with data from large numbers ofother patients with similar disease manifestations.What Is Artificial Intelligence?AI has been defined as an area of study in computer scienceconcerned with the development of computers to engage inhuman-like thought processes such as learning, reasoning andself correction [6]. The phrase artificial intelligence isbelieved to have been used first at a Dartmouth CollegeConference in 1956 [7]. AI allows programmers and usersto overcome the many constraints of traditional decision sup-port approaches, such as rule-based systems, which includedifficulty in rule formulation and challenges in updating newrules. These traditional systems, although created with expertinput, do not exhibit human behaviors, such as reasoning, self-improvement, and constant learning. Despite extensive effortsand initial excitement, the application of AI has fallen short ofits potential in medical applications. However, AI has experi-enced something of a renaissance within the past few years innonmedical applications [7].For example, Siri, which was originally introduced as anApple (Apple, Inc., Cupertino, CA) iOS application by Siri,Inc., has become embedded as a major feature of iPhones,starting with the 4S (as well as the third-generation iPad), andhas become one of the most popular AI applications andarguably, the best feature of the iPhone today. Siri serves asan intelligent personal assistant that can provide schedulingand information, such as time, weather, local restaurant facts,or directions, by connecting to the vast array of data availableon the Internet.By using a combination of speech recognition, naturallanguage processing, and AI, Siri performs relatively mun-dane tasks that humans can do, such as look at a map or askanother person for directions, and for the most part, under-stands commands and performs with a minimum of errors.How does Siri limit mistakes? Directional navigation providesone example. When a person asks another person for direc-tions, there is always a chance that the results could bemisleading or incorrect. Because Siri is connected to theInternet, the application can access correct information andaccurately direct the individual to the requested destination.Siri can also provide step-by-step support on the optimal routeto reach the destination as well as time and distance required.Although Siri is remarkable, its ability to respond to diag-nostic or therapeutic medical questions is limited to its abilityto initiate a search on the Internet. However, its bigger brother,Watson, has begun to be used in health care applications.IBMs Watson, best known for its remarkable performanceon Jeopardy! , provides many unique and transformative pos-sibilities to resolve challenges associated with medical diag-nosis and treatment. The Watson hardware, which costs ap-proximately /3million, can process 500 gigabytes/second, theequivalent of 1 million books [8]. Much as Siri helps guideusers to the best route to desired destinations, AI applicationsin medicine, such as Watson, may help physicians navigatethrough a complex set of patient symptoms, laboratory data,and imaging results to come up with a set of most likelyclinical diagnoses and treatment options that may ultimatelyimprove patient outcomes and reduce health care costs. IBMinitially tested its developing AI medical acumen using theAmerican College of Physicians Medical Knowledge SelfAssessment study guide and subsequently improved on itsperformance by adding textbooks, such as the Merck Manualof Diagnosis and Therapy and additional medical journals andbooks that were not included in the original Jeopardy! data-base. The IBM developers then improved on the performancefor medical applications by fine-tuning the weighting associ-ated with the various algorithms utilized by the application for441, Page 2 of 8 Curr Cardiol Rep (2014) 16:441these medical domain questions. The team has also created ademonstration in which a patients presenting symptoms areinput into the software and a series of progressive questionsare posed by a health care worker to personalize the diagnosticand therapeutic recommendations made by the software. Twoimportant features of the prototype software were the ability toprovide multiple possible diagnoses and treatment optionswith relative confidence levels and the ability to trace theinformation utilized to make a recommendation.Potential Applications of Artificial Intelligencein MedicineOne promising initial application of such technologies is totake advantage of the relative computational speed availablein todays computers made possible by parallel processing.This speed allows performance of on-the-fly syntheses of apatients electronic medical record from one or more sourcesand creation of a summary and current problem list. Patientproblem lists are typically poorly organized and managed,with no single provider given overall responsibility. The po-tential to create a graphical synthesis of patient data using acombination of natural language processing and AI technolo-gy is exciting. Not only can AI systems perform a rapid andthorough search of single or multiple patient electronic med-ical records, such systems can also search the Internet, text-books, and journals for data. IBMs Watson, when integratedinto practice, could potentially search through its exhaustivedatabase, which would be kept up to date with current litera-ture, to support accurate diagnoses and provide other possiblediagnostic and treatment options. This technology could beutilized to cross correlate data from a patients family history,find patients similar to that patient, and evaluate ultimatediagnoses and treatment responses. As genomic, proteomic,and metabolomic databases become commonplace andsearchable, the software will be able to utilize these data inmaking recommendations for patient screening and in formu-lating diagnostic and treatment recommendations. In additionto providing answers, the software could be utilized to askadditional pertinent questions to more effectively and safelydirect a diagnostic work-up plan and the performance of teststhat maximize efficacy and safety while minimizing healthcare costs.Physician time with patients is currently limited. The 1995Commonwealth Fund survey of physicians found that 29% ofphysicians were dissatisfied with the amount of time theyspent with patients and only 31 % were very satisfied. Inaddition, 41 % reported a decline in time with patients be-tween 1992 and 1995 [9]. Recent cuts in reimbursement haveput additional pressure on physicians with regard to time spentwith patients. Limited time to spend with patients may con-tribute to rising errors and incorrect diagnoses. In 1993, theaverage visit length was 20 minutes for family practitionersand 26 minutes for general internists [9]. Today the averagevisit length for family practitioners has dwindled to10 minutes, a timespan that severely restricts the ability toobtain an adequate understanding of patient symptoms, makean accurate diagnosis, and provide thoughtful care. Giventodays increasing patient loads and requirements for docu-mentation, including interfacing with electronic health re-cords, it seems inevitable that AI applications will be widelydeployed in the next few years. As is true of other areas ofhealth care information technology, advances and productsdesigned outside of the medical space can be as many as1015 years ahead of practical application in health care. Byutilizing these advanced computing technologies, physiciansof the future may spend less time behind a computer and moretime with patients, reversing the current trend.The cost of health care will be another major driver in theapplication of AI in medicine. AI applications will be utilizedto reduce unnecessary testing, decrease the disparity and dis-crepancies in care throughout the United States and the rest ofthe world, and reduce hospital admissions and length of stay.Current Application of Artificial Intelligence in CardiacImagingAI can be used in the field of cardiology in a number of waysas shown in Fig. 1, including determining the most appropri-ate type of imaging study for a specific set of symptoms [10].If applied properly, AI could reduce inappropriate imagingstudies and help physicians adhere to practice guidelines andever-changing appropriate use criteria. For example, theImaging in FOCUS (Formation of Optimal CardiovascularUtilization Strategies) quality improvement initiative of theAmerican College of Cardiology was recently introduced toreduce inappropriate use of diagnostic imaging through theuse of AI that tracks appropriate use criteria [11]. Among 55participating sites that voluntarily completed the radionuclideimaging performance improvement module, the proportion ofinappropriate cases decreased from 10 % to 5 %. Thesepreliminary data from initial participating sites suggest thatthrough the use of self-directed, quality improvement soft-ware, and an interactive community, physicians may be ableto significantly decrease the proportion of tests not meetingappropriate use criteria [11].After images are acquired, additional AI tools may helpphysicians provide accurate interpretation of cardiac imagingstudies [12]. Artificial neural networks are an example of waysin which AI systems that approximate the operation of thebrain can be successfully applied in cardiac imaging [13].These networks have been utilized in diagnosis and treatmentof coronary artery disease and myocardial infarction, interpre-tation of electrocardiographic studies, detection of arrhythmiasCurr Cardiol Rep (2014) 16:441 Page 3 of 8, 441such as ventricular fibrillation [14], and in image analysis forechocardiography and other cardiac imaging, as well asscreening for heart murmurs in children [15].Artificial neural network technology is complex and in-volves multiple processors working in parallel. Neural net-works can be supervised (human feedback on the data andanalysis) or unsupervised. A neural network is initiallytrained with large amounts of data and rules about relation-ships among those data elements. In what are known asfeedforward systems, learned relationships can then be uti-lized to inform higher layers of knowledge. These networksutilize multiple approaches, including gradient-based training(adjustment of the weights of various elements to minimizethe gradient of error), fuzzy logic, Bayesian methods, andFig. 1 Flowchart of artificialintelligence applications in apatient with suspected myocardialischemia441, Page 4 of 8 Curr Cardiol Rep (2014) 16:441genetic algorithms. In general, neural networks use weightingfactors (depending on the importance of the category) andattempt many different combinations of weightings until themost accurate answer is identified.Nuclear imaging is evolving from subjective (moreart than a science) approaches to more objective, digital-based quantitative techniques, providing insight into the phys-iologic processes of cardiovascular disorders and predictingpatient outcomes [16]. The digital-based nature of nuclearimages permit the application of automated quantitative soft-ware to assist in interpretation of cardiac single-photon emis-sion computed tomography (SPECT) and positron emissiontomography (PET) studies. Using AI approaches, algorithmshave been developed that take raw digital data output bySPECT and PET cameras, identify the location of the heart,and reconstruct tomographic slices of the heart into 3-dimensional sets of transaxial sections, all automated withoutoperator interaction. These automated image analysis systemsthen analyze signals from several hundred regions of the heartin transaxial sections and compare the intensity of these sig-nals with those expected in a normal sex- and radiotracer-matched heart to generate a quantitative map of the location,extent, and severity of regional signal differences. Dependingon the radiotracer used, this approach provides objectiveinformation on myocardial perfusion, metabolism, and/or in-nervation and aids physicians with both the interpretation ofimages as well as assistance in selection of optimal treatmentstrategies. Beyond static dataset evaluation, dynamic mea-surement of the heart cavity andmyocardium can be evaluatedusing electrocardiographically gated 3-dimensional images byautomatically identifying the endocardial and epicardial sur-faces of the left ventricular cavity and following their motion(contraction and thickening) throughout the cardiac cycle [17].In SPECT and PET imaging, AI approaches can also beemployed to highlight an abnormality and thus serve in anadjunctive capacity rather than suggesting a primary diagno-sis. In this use case, AI serves as an image enhancementtechnique rather than as a diagnostic tool. This approach canresult in an increase in both accuracy and speed of interpreta-tion. In another application, this decision support technologycan help to determine the limits of normal in a specific patientpopulation. This is particularly useful given the variation innormal distribution of a specific radiotracer, (which is appliedin order to mirror the distribution of blood flow in the heart) innormal subjects. For example, what would be considered torepresent normal physiologic distribution of 13N-ammonia asa blood flow agent in the lateral region of the left ventricularmyocardium may, in fact, represent an abnormal lateral per-fusion defect with other myocardial perfusion radiotracers,such as 201Tl [18]. Although the interpreter may not be ableto discern such subtle differences between cameras and radio-tracers, AI applications can be utilized to access a large normaldatabase to aid the physician in rendering the correctdiagnosis. Thus, much like Siris ability to accurately directan individual to the requested destination (by identifying thecurrent location of the individual and comparing it to a largeGoogle Maps database), AI applications in cardiac imagingare and will be used to help the physician to correctly interpretcardiac images by comparing the patients tomographic im-ages with a large age- and sex-matched normal database that isspecific for the radiotracer and the camera used in acquiringthe patients images.Future Application of Artificial Intelligence in CardiacImaging: Neural Networks and WatsonThe use of AI systems, such as Watson, represent a novelarchitecture for evaluation of unstructured and structured con-tent when compared with traditional expert systems thatused forward reasoning (data to conclusions) or backwardreasoning approaches (such as the if-then statements usedby Stanford Universitys MYCIN system and others thatfollowed). These previous expert systems were costly, difficultto develop and maintain, and were brittle, requiring a perfectmatch between input data and existing rule forms. Rule formsare inherently limited in assumptions about which questionswill be posed. Software such as that utilized byWatson, on theother hand, uses natural language processing and a variety ofsearch techniques to create hypothesesmaking it more flexible,scalable, easy to maintain, and cost effective. This new ap-proach makes it much easier to keep up with ever-changinginformation in imaging, medicine, and surgery.In a future clinical image interpretation scenario utilizingAI technology, a requested study would first be evaluated forappropriateness based on the patient history, previous exam-inations and their interpretations, and the indication for theexamination. The examination would, if deemed appropriate,then be protocoledwith regard to the way in which it shouldbe performed, amount and type of radiopharmaceutical and/orcontrast material, MR imaging sequences, etc. to optimizeefficacy, safety, and efficiency.As a next-generation AI system assists in interpretation of anuclear scan, for example, it would consider many variablesand assign different weighting factors in order of their impor-tance, such as whether the patient has had a previous myocar-dial infarction in the targeted area, risk factors for coronaryartery disease, prior coronary angiography, percutaneous orsurgical revascularization, and medical therapy based on em-pirical experience using big data mining techniques. All ofthese would be weighted and, because the computer wouldtake them all into account, it might come up with multiplepossible candidate diagnoses within a short period of time,with associated probabilities that might improve image reviewaccuracy and efficiency. The software would also act as an aidin diagnosis by pointing out certain areas in a scan that shouldCurr Cardiol Rep (2014) 16:441 Page 5 of 8, 441be reviewed carefully because they fall outside expected nor-mal parameters personalized for a particular patient. As is truein human learning, the AI systems of the future will becomebetter over time because they will have access to large num-bers of imaging studies and their results. Prior recommenda-tions and subsequent outcomes will be iteratively fed backinto the technologys algorithms, resulting in improved accu-racy and efficacy.Barriers to Widespread AI Use in the Near FutureIn addition to its tremendous potential and recent advances, AItechnology faces many obstacles before it can be fully imple-mented in medicine as outlined in Table 1. It must be suffi-ciently fast to be accepted by users and must be integrated intophysician workflow, which will mean that it must be tightlyinterfaced with electronic medical record systems and, in thecase of medical imaging, into image interpretation workflowat image review workstations.Another challenge is in determining the accuracy of thesystem for diagnoses and treatment recommendations andother expert system applications. Multiple opinions from phy-sicians in different specialties and subspecialties often conflictin a single case (as is seen in many malpractice cases). Thegold standard in medical practice is not always obvious; noris there consensus on whether such a standard should reflectexpert opinion, the majority opinion among physicians, orbest reported outcomes in similar settings.Hospital and outpatient practices, encouraged by the Federalgovernments meaningful use initiative, have been makingmajor investments in electronic medical record systems. It isnot clear what level of accuracy would be required for hospitalsto make additional investments in AI systems, even with evi-dence indicating added value in a clinical setting. It is unlikelythat the Centers for Medicare & Medicaid Services or otherpayers will provide additional reimbursement for AI technolo-gies in the near future.Another major barrier comes from the regulatory and med-icolegal perspectives. Although society is increasingly accus-tomed to information technologyrelated glitches and prob-lems, such technical errors and failures are likely to be deemedunacceptable in medical and health applications. Consumersare already aware that many onlineWeb sites provide unreliablediagnostic information, incorrect pharmaceutical dosages, inac-curate medication expiration data, and more. Most vendors thathave developed software in the medical space have avoidedsystems that render medical diagnoses or treatment recommen-dations because of perceived liability and the real difficultiesassociated with obtaining U.S. Food and Drug Administrationclearance for this type of software.Another major challenge involves concerns about patientprivacy and security. Federal regulations place growing re-strictions on private health data, but no matter how impene-trable computer programs and software are required to be, therisk of privacy breaches is ever present. An additional chal-lenge lies in the competitive nature of the information tech-nology market and the reluctance of electronic medical recordvendors to provide a highly integrated solution to a third-partyprovider of AI software.The use of AI also raises specific medicolegal concerns.For example, if a physician uses an AI system to help inTable 1 Opportunities and challenges for fully implementing artificial intelligence in medicineOpportunities ChallengesThe transition from a paper based to an electronic medical recordmakes rapid digital record searching possibleLack of access to large de-identified databases such as multihospital ormultinetwork EHRsSubstantial advances in computer technology including multi-coreprocessing and new approaches such as IBMs Deep Q/A thatallow the software to learn and stay current with the literatureAlthough there is some precedent for CAD reimbursement formammography, it is unclear how healthcare will pay for these systemsgiven the lack of direct reimbursement for AI systems in medicineDiagnostic errors in medicine are surprisingly common and mostlyrelated to cognitive factors. Computer algorithms do not makethe same type of mistakes as humans; consequently they can becomplementary to and synergistic with human diagnosesPerceived medicolegal and regulatory challenges such as liability formisdiagnosis or incorrect treatment recommendations and difficultyproving improved efficacy of AI systemsWidespread use of an AI system could reduce the well documenteddisparities and discrepancies in patient diagnosis in differentregionsDifficulty in determining a gold standard for accuracy in diagnosis of acomputer system given the variability in expert diagnosis in humansArtificial intelligence systems can be used for teaching, especiallywhen a practitioner drills down to review the programs logicConcerns about privacy and security have resulted in limitations in accessto patient dataIncreasing availability, sophistication and trust in computer expertsystems in society in generalDifficulties in accessing patient data in clinical trials for clinical use and intranslating clinical guidelines into a machine readable formatThe potential for improved productivity/efficiency as well asimproved accuracyThe potential for an AI clinical decision support system to result in adisruption or slowdown in workflow and productivityAI artificial intelligence, EHR electronic health records, CAD computer aided detection441, Page 6 of 8 Curr Cardiol Rep (2014) 16:441clinical diagnosis and provide best treatment options for apatient, what happens when diagnosis turns out to be incor-rect? If misdiagnosis resulted in delayed or incorrect treat-ment, then who should be medically liable for the adverseoutcome? Should it be the authors of the software, the tech-nology provider, the hospital who provided the technology,the doctoror all of the above? Conversely, what are themedicolegal implications of not following an AI systemsrecommendations with a subsequent adverse result? Theseand other perceived liability questions might result in softwaredevelopers and medical providers perceiving AI systems as amedical liability. It is unclear what the drivers will be inreaching a point at which the perceived benefits of such asystem outweigh potential downsides.ConclusionsThe practice of medicine is at a crossroads, with simultaneousincreases in patient volume, an explosion in the amount andcomplexity of medical and scientific knowledge, and thetransition to electronic medical records. This is occurring atthe same time that the ratio of physicians and other health careproviders to patients continues to decrease andwhile cognitivedemands on physicians are at an all-time high. As theSeptember 2013 Institute of Medicine report on the nationscancer care system stated, the United States has an increas-ingly chaotic and costly medical system that is in crisis andfails to deliver consistent care that is evidence based, coordi-nated, and patient centered [19]. The combination of AI, bigdata, and massively parallel computing offers the potential tocreate a revolutionary way of practicing evidence-based, cost-effective, and personalized medicine. However, barriers toadoption of AI technologies must be overcome from regula-tory, legal, cultural, and political perspectiveseven whentechnology solutions have matured. Cardiac imaging has beena relatively early adopter of AI techniques in image process-ing, structured reporting, and clinical decision support systemsand can continue to lead the way for the rest of medicalimaging and the practice of medicine.Acknowledgment The authors wish to acknowledge and thank NancyKnight for her tremendous assistance in the editing of this manuscript andStephen Siegel for his assistance with the graphics.Compliance with Ethics GuidelinesConflict of Interest Steven E. Dilsizian declares that he has no conflictof interest. Eliot L. Siegel has received PI funding for a grant from IBM tohelp bring the Jeopardy! Deep Q/A software to the medical domain.Human and Animal Rights and Informed Consent This article doesnot contain any studies with human or animal subjects performed by anyof the authors.ReferencesPapers of particular interest, published recently, have beenhighlighted as: Of importance Of major importance1. LeeCS, Nagy PG,Weaver SJ, Newman-Token DE. Cognitive andsystem factors contributing to diagnostic errors in radiology. Am JRoentgenol. 2013;201:6117. This article provides an excellentoverview of the types of diagnostic errors made by radiologists inthe interpretation of imaging studies including common biases .2. Graber ML, Franklin N, Gordon R. Diagnostic error in internalmedicine. Arch Intern Med. 2005;165:14939.3. Weingart SN,Wilson RM, Gibbard RW, Harrison B. Epidemiologyof medical error. West J Med. 2000;172:3903.4. Khosla V. 20 % doctor included: speculations and musings of atechnology optimist. Available at: http://www.khoslaventures.com/wp-content/uploads/2012/12/20_percent_doctor_included.doc.Accessed October 14, 2013.5. Winters B, Custer J, Calvagno Jr SM, et al. Diagnostic errors in theintensive care unit: a systematic review of autopsy studies. BMJQual Saf. 2012;21:894902.6. Smith SS, Morris W, Voorheis RW. The new internationalWebsters comprehensive dictionary of the English language: en-cyclopedic edition. Naples: Trident Press International; 1998.7. Patel VL, Shortliffe EH, Stefanelli M, et al. The coming age ofartificial intelligence in medicine. Artif Intell Med. 2009;46:517.This paper provides an overview of a panel discussion at theArtificial Intelligence in Medicine Europe conference in July 2007focusing on predictions of the coming of age of AI in the 1990s thatdid not materialize and research and experience during the subse-quent years. Topics included clinical decision support, dealing withuncertainty, systems integration, and modeling of expertise .8. Pearson T. IBM Watson. How to build your own Watson Jr. inyour basement. IBM Developer Works blog. February 18, 2011.Avai lable at : ht tps: / /www.ibm.com/developerworks/mydeveloperworks/blogs/InsideSystemStorage/entry/ibm_watson_how_to_build_your_own_watson_jr_in_your_basement7?lang=en. Accessed September 30, 2013.9. Dugdale DC, Epstein R, Pantilat SZ. Time and the patient-physician relationship. J Gen InternMed. 1999;14 Suppl 1:S3440.10. Fornell D. An introduction to clinical decision support for cardiol-ogy. Diagn Intervent Cardiol. 2013. Available at: www.dicardiology.com/article/introduction-clinical-decision-support-cardiology. Accessed September 29, 2013.11. Saifi S, Taylor AJ, Allen H, Hendel R. The use of a learningcommunity and online evaluation of utilization for SPECT myocar-dial perfusion imaging. J Am Coll Cardiol Imaging. 2013;6:8239.12. DePuey GE, Garcia EV, Ezquerra NF. Three-dimensional tech-niques and artificial intelligence in thallium-201 cardiac imagingcardiac imaging. Am J Roentgenol. 1989;152:11618.13. Itchhaporia D, Snow PB, Almassy RJ, Oetgen WJ. Artificialneural networks: current status in cardiovascular medicine. J AmColl Cardiol. 1996;28:51521. The authors define artificial neuralnetworks as a way to find unexpected relations in data sets based oninput from previous data and describe research performed in the1980s and 1990s in cardiovascular medical research.14. Clayton RH,Murray A, Campbell RWF. Recognition of ventricularfibrillation using neural networks. Med Biol Eng Comput. 1994;32:21720.15. DeGroff CG, Bhatikar S, Hertzberg J, Shandas R, Valdes-Cruz L,Mahajan RL. Artificial neural networkbased method of screeningheart murmurs in children. Circulation. 2001;103:27116.Curr Cardiol Rep (2014) 16:441 Page 7 of 8, 44116. Dilsizian V, Taillefer R. Journey in evolution of nuclear cardiology:will there be another quantum leap with the F-18 labeled myocardialperfusion tracers? JACC Cardiovasc Imaging. 2012;5:126984.17. Artificial Intelligence in Medicine. (n.d.). Retrieved September 22,2013, from Cedars Sinai: http://www.cedars-sinai.edu/Patients/Programs-and-Services/Medicine-Department/Artificial-Intelligence-in-Medicine-AIM/.18. Dilsizian V, Bacharach SL, Beanlands RS, et al. ASNC imagingguidelines for nuclear cardiology procedures: PET myocardialperfusion and metabolism clinical imaging. J Nucl Cardiol.2009;16: published online as supplementary material. Availableat: http://link.springer.com/content/esm/art:10.1007/s12350-009-9094-9/file/MediaObjects/12350_2009_9094_MOESM1_ESM.pdf. Accessed September 30, 2013.19. Levit L, Balogh E, Nass S, Ganz PA, editors. Deliveringhigh-quality cancer care: charting a new course for a systemin crisis. Institute of Medicine. Washington, DC: Institute ofMedicine; 2013.441, Page 8 of 8 Curr Cardiol Rep (2014) 16:441Artificial...AbstractIntroductionWhat Is Artificial Intelligence?Potential Applications of Artificial Intelligence in MedicineCurrent Application of Artificial Intelligence in Cardiac ImagingFuture Application of Artificial Intelligence in Cardiac Imaging: Neural Networks and WatsonBarriers to Widespread AI Use in the Near FutureConclusionsReferencesPapers of particular interest, published recently, have been highlighted as: Of importance Of major importance

Recommended

View more >