Techniques in Evaluating Nursing Expert Systems: A Case Study

This study addresses the problems in evaluating nursing diagnostic artificial intelligence (AI) expert systems. Two separate experiments (N=49) were conducted using a computer expert system. The first experiment, the "white box" experiment (n=9), compared the diagnostic techniques applied by experienced RNs against the programmed techniques used by the expert system. The second experiment, the "black box" experiment (n= 40), compared diagnostic results of beginning nurses against the computer expert systems results. In some cases the computer outperformed the nurses and vice versa. The evaluation techniques, as applied in both experiments, enhance the ability of nurses to evaluate and select AI expert systems to be used in computer-assisted diagnosis of nursing problems.     

Interpretation of cutaneous biopsy specimens: choice of pathologist by primary care practitioners

BACKGROUND: Primary care providers (PCPs) have limited training in recognizing common skin disorders, and additional emphasis may be placed on laboratory evaluation, including skin biopsies. METHODS: Primary care providers in Miami, Fla, were surveyed regarding skin biopsy and excision practices and histologic interpretation. They then participated in an examination, using 20 high-quality color photographs of common dermatologic disorders. RESULTS: Of the 80 PCPs who participated, 42% currently do skin biopsies and consult general pathologists for interpretation. Another 20% of PCPs intend to do biopsies within the next 5 years. Only 33% prefer to have a dermatopathologist interpret histopathology. We found no correlation between PCPs' scores on diagnostic testing and whether they do skin biopsies. CONCLUSION: Nearly all PCPs doing skin biopsies in our study sample use general pathologists to interpret histopathology. The limited training in clinical dermatology of both PCPs and general pathologists may result in compromised clinical-pathologic correlation.     

Understanding laboratory test results. Conditions for appropriate use of laboratory tests

The appropriate use of laboratory tests requires that valid, reliable, and reproducible data be obtained and that the clinician know both how to interpret the information provided by diagnostic tests and how to apply it to individual clinical situations. Appropriate interpretation and clinical use of diagnostic tests requires that clinicians understand the principles of laboratory testing, the information provided and not provided by laboratory tests, and how to evaluate the clinical efficacy of laboratory tests. As laboratory testing moves from regional and hospital laboratories into office laboratories, clinicians need to become more knowledgeable regarding the technical and quality control factors that affect diagnostic test accuracy. This will require the establishment of stronger, more effective links between clinicians and laboratory pathologists.     

Automatic evaluation of arterial diameter variation from vascular echographic images

An automatic procedure for accurate arterial diameter evaluation from B-mode images obtained by diagnostic ultrasound systems is presented. It is used for measuring brachial artery dilation following reactive hyperemia induced by forearm ischemia, which is an appropriate parameter to study endothelial function in humans. B-mode images obtained from the diagnostic ultrasound system are acquired on a personal computer as grey intensity fields (pixels). A completely automatic algorithm is then applied and the artery walls are identified by two discrete sets of points. Artery diameter is evaluated by parabolic least-square approximation. The accuracy and extensive range of applicability of the diameter evaluation procedure were demonstrated both by preliminary analytic test cases and in vivo analyses. Reproducibility of the diameter estimate was assessed by in vivo measurements. The proposed procedure permits fast and accurate analysis of large amounts of data, because it requires no action by the operator. It thus represents a valuable tool for assessment of endothelium-dependent vasodilation, especially in large, multicentric clinical trials. (E-mail: l.ghiadoni@med.unipi.it)     

人工智能在病理诊断领域的进展

基于近几年机器视觉的发展,深度学习的人工智能方法应用于组织病理极大程度上促进了病理学家解决临床上的诊断问题,用该种方法解决病理学问题可被称为计算机病理学。人工智能可以做到帮助病理学家初筛大部分良性数据、辅助诊断、疗效预测、识别生物标志物等,甚至可以做到对药效治疗监测以及识别药物发现未知的信号。基于深度学习在病理领域的深入研究,让计算机自动处理病理数据成为可能。人工智能诊断决策建立在大数据之上,很多有可能做到对每个病人的个性化管理,对于大多普遍性的疾病诊断有着更加快速准确的优势。但数字病理学的发展仍受到一些问题的限制,以至于现阶段没有广泛应用于数字病理诊断平台。本文总结了近几年人工智能在病理诊断领域的最新进展,并讨论这种技术的可行性,补充说明在数字病理学中遇到的困难和挑战,并提出在该领域实用性上的展望。     

Laboratory clinical effectiveness: pathologists improving clinical outcomes

This article calls on pathologists to take a larger role in improving the performance of the American health care system. To improve outcomes for populations and individuals require that pathologists increase their activities outside of the traditional laboratory in interdisciplinary collaborations, outcomes research, health care systems development, and clinical care.     

TREACT. An expert system consultation program to aid in the diagnosis of transfusion reactions

We have developed a rule-based, expert system consultation program, TREACT, to aid in the diagnosis of transfusion reactions. Given clinical signs, symptoms, and laboratory results, the program generates diagnoses, alerts the user when a medical director should be called, suggests follow-up actions, and makes recommendations for future transfusion. Diagnoses made by TREACT, including 121 reactions, were compared with those of the medical directors over a 6-month period. The overall diagnostic concordance between the medical directors and the program was 0.777. When this was corrected for chance association (kappa statistic), the concordance was 0.703 (p less than 0.0001), which can be interpreted on a qualitative scale as substantial agreement. The program has also been used successfully as a tool for training new technologists. Other advantages and possibilities that expert systems offer to transfusion medicine are discussed.     

Impact of laboratory testing on DRG coding and reimbursement--results of a data base research

We filed an Australian data base with about 7 million well-documented clinical cases for comorbidities, which can be proven, supported or excluded by laboratory testing and which have an impact on DRG reimbursement. The result was a list of 123 DRGs being shifted to a higher severity level by documenting one out of 157 complications. For better visualization of the more than 4,000 combinations, we developed a computer program, which allows the laboratory to develop its own diagnostic pathways for these complications in a simple Excel format.     

HDL cholesterol as a sensitive diagnostic parameter in malaria]

In patients with malaria the lipid parameters triglycerides, cholesterol, and HDL-cholesterol were determined routinely. At the time of admission hypertriglyceridemia, hypocholesterolemia, and an extreme decrease in HDL-cholesterol were found. This dyslipoproteinemia was present in cases of falciparum malaria, as well as in cases of benign tertian malaria. The extent of HDL-cholesterol decrease showed no correlation to the severity of the clinical course of disease. HDL-cholesterol has proven to be an independent diagnostic laboratory finding in cases of suspected malarial infection. This parameter displays high diagnostic sensitivity, but no specificity for malaria.     

Improving the utilization of clinical laboratory tests

Reimbursement policies for health care services are greatly diminishing in the U.S. and Western Europe. Hence, there is an increasing need for doctors and other care givers to reduce costs without compromising the quality of the care being delivered. The clinical laboratory is viewed as an area of high costs where significant reductions have been targeted. Efficient utilization of laboratory services can be achieved by elimination of the general health panel, removal of old tests or those that provide redundant information, a reduction in the use of standing orders, more judicious use of drug assays, acceptance of clinical practice guidelines, and use of reflex testing algorithms. New technologies such as DNA probes can substantially improve diagnostic efficiency. Point-of-care testing devices which have higher costs than incremental central laboratory expenses should only be used if they reduce overall operating expenses. Implementation of expert systems can make remaining tests more effective. Doctors and laboratorians must collaborate to achieve more efficient utilization practices.     

Expert nurses and expert systems. Research and development issues

Expert systems in nursing have developed primarily through deductive, theoretical approaches and have, for the most part, failed to articulate the need for expert clinical nurse knowledge and heuristics as a basis for developing expert systems. Research using knowledge engineering and nurse experts, as well as multi-method approaches to researching nurse experts, should provide content for development of computer systems, in general, and expert systems, in particular, in nursing.     

Laboratory computers: tools for increased productivity

Revolutionary changes in the computer industry brought about in part by the introduction of personal computers are now reaching into clinical laboratories everywhere. Although the true justification for any computer tool may be an intangible one such as improvement in service, such tools can typically be cost justified by increases in productivity alone. A broad spectrum of applications software useful in the medical laboratory is now available in any scale required to meet the needs of any size of laboratory. Perhaps one of the most dramatic changes in this field in the last several years has been the introduction of small-scale integrated laboratory systems that are true small-scale laboratory information systems. As a result, smaller laboratories can now acquire laboratory information systems appropriately scaled to their workload for a fraction of the cost of the large systems. These small laboratory information systems are typically more easily cost justified than larger ones. Achieving productivity gains using computer tools in the laboratory is a management-intensive process requiring careful analysis and thorough planning. Laboratory managers who eschew computer tools are now an anachronism; extinction of this species is imminent.     

The use of knowledge-based systems to improve medical knowledge about urine analysis

Urine protein diagnostics has developed into a routine method for screening and monitoring kidney diseases. It is based on the quantitative measurement of total protein, albumin, alpha(1)-microglobulin, immunoglobulin G and alpha(2)-macroglobulin (all related to urine creatinine), as well as a dipstick screening. The excretion pattern of the marker proteins allows differentiation of haematuria, leukocyturia and proteinuria and to assign them to prerenal, renal and postrenal causes. In order to provide the clinical partner not only with pure analytical results, but to support clinical decision making by an interpretative report, a urine protein expert system (UPES) has been developed. Based on a database containing more than 500 excretion patterns of patients with known diagnoses, a knowledge base was extracted. In its modules plausibility control, glomerular filtration rate, hematuria, leukocyturia and proteinuria, IF-THEN-rules interpret the given patterns and select matching text elements. The knowledge base has been integrated in the modern expert system shell WILAS, and the resulting interpretation system has been thoroughly verified and validated. An internal acceptance study revealed that urine protein differentiation is widely accepted as a diagnostic option and that its interpretation, provided with the help of UPES, is appreciated as a service. In an external study, the usability of UPES in routine and its knowledge representation was evaluated in 11 centres consisting of laboratories and nephrological partners. Over seven months, more than 500 cases were interpreted using UPES and documented by questionnaires. The discussion of the results at a user conference revealed that the problem of analytical standardisation as well as the common definition of diagnostic terms by laboratory staff and clinicians play a crucial role for the use of knowledge-based systems in laboratory medicine. Whereas the user interface of UPES was judged very heterogeneously, the correctness of the proposed interpretations was unanimously rated as "good". As a result of the evaluation, the user interface has been modernised. The knowledge base has been extended to address paediatric issues as well, and to take clinical information and previous findings into consideration.     

The application of expert systems in the clinical laboratory

An "expert system" consists of a knowledge base containing information of a general nature and an inference system that receives data from the user and applies the knowledge base to produce advice and explanations. An expert system stripped of its knowledge base (a tool) may be used to build new expert systems. Existing systems relevant for laboratory medicine are reviewed. The role in the laboratory of expert systems and their integration and evaluation are discussed.     

Neuroimmunologic parameters in therapy refractory epilepsy in children

In a prospective study comprising 48 children (21 boys, 27 girls, age 12 months to 15 years) the hypothesis was tested whether an alteration in the light chain immunoglobulins kappa (kappa) and lambda (lambda) is an expression of autoimmune mechanisms and could be used as a diagnostic parameter. Immunoglobulins were analysed by kinetic nephelometry. The patients were grouped according to clinical and laboratory parameters. A high kappa/lambda ratio caused by high concentration of the kappa chain was found in children with therapy-resistant epilepsy. 60% showed a high kappa/lambda ratio, no distinct association was ascertainable in 26% cases and in 14% of the cases false positive results were obtained. In seizure-free children immunological markers were specific in 72% of cases, whilst 24% could not be characterized and 4% showed false positive results. The kappa/lambda ratio, therefore, is a good diagnostic aid in the treatment of therapy-resistant epilepsies.     

Statistical techniques for evaluating the diagnostic utility of laboratory tests.

Clinical laboratory data is used to help classify patients into diagnostic disease categories so that appropriate therapy may be implemented and prognosis estimated. Unfortunately, the process of correctly classifying patients with respect to disease status is often difficult. Patients may have several concurrent disease processes and the clinical signs and symptoms of many diseases lack specificity. In addition, results of laboratory tests and other diagnostic procedures from healthy and diseased individuals often overlap. Finally, advances in computer technology and laboratory automation have resulted in an extraordinary increase in the amount of information produced by the clinical laboratory; information which must be correctly evaluated and acted upon so that appropriate treatment and additional testing, if necessary, can be implemented. Clinical informatics refers to a broad array of statistical methods used for the evaluation and management of diagnostic information necessary for appropriate patient care. Within the realm of clinical chemistry, clinical informatics may be used to indicate the acquisition, evaluation, representation and interpretation of clinical chemistry data. This review discusses some of the techniques that should be used for the evaluation of the diagnostic utility of clinical laboratory data. The major topics to be covered include probabilistic approaches to data evaluation, and information theory. The latter topic will be discussed in some detail because it introduces important concepts useful in providing for cost-effective, quality patient care. In addition, an example illustrating how the informational value of diagnostic tests can be determined is shown.     

Computer-aided simple triage

Purpose

Computer-aided detection (CAD) established its role in medical imaging as second reader aimed to boost the diagnostic accuracy of human interpreter. As the diagnostic performance of CAD systems improves and more imaging modalities are covered, CAD steps forward to fill new, more demanding positions in medical practice. In this paper, we investigate how the introduction of CAD for emergency diagnostic imaging shifts the use case paradigm from second reader to initial interpreter and triage tool.

Methods

We start from extracting common characteristics of exiting CAD systems and compare them to those for emergency diagnostic imaging modalities. Based on the deduced requirements, we define a new class of CAD systems??Computer-aided simple triage (CAST) and explore its properties, use case scenarios and clinical benefits. We also discuss the differences between the CAST, CAD, and automated computer diagnosis.

Results

A CAST system should serve as a simple triage tool performing a fully automatic analysis and providing initial classification at ??per study?? level. Positive studies are then immediately analyzed by expert reader, thus reducing delay for patients with critical conditions, while negative studies can be initially dealt with by less experienced staff. Automatic image quality and study complexity assessment can serve as reading prioritization key. CAST system should exhibit sufficiently high specificity, while not compromising the high sensitivity per study.

Conclusions

CAST systems have a potential to become an ??enabling technology?? allowing introduction of advanced imaging techniques into the emergency workflow protocols by addressing the reader unavailability and reading prioritization problems.     

Using the laboratory information system to achieve strategic advantage over the competitors of hospital-based clinical laboratories

The competitors of hospital pathologists are commercial reference laboratories, hospital special function laboratories, pathology groups in neighboring hospitals, hospital mainframe computer personnel, and users of patient proximity testing systems. The laboratory information system can provide a strategic advantage over these competitors by matching, exceeding, or substituting for their capabilities and by creating switching costs for clinics, administrations, and patients. For example, the installation of microcomputers in clinicians' private offices provides them with ready access to the pathology data base, bonds them to the hospital, and capitalizes on the willingness of the hospital to invest in information technology.