SEARCH · Search PubMed
Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Data mining is a recent development in computer science that combines artificial intelligence algorithms and relational databases to discover patterns automatically, without the use of traditional statistical methods. Work with data mining tools in health care is in a developmental stage that holds great promise, given the combination of demographic and diagnostic information.
When a person sizes up a complex situation and rapidly comes to a decision that proves to be not just good but brilliant, you think, "That was smart." After you watch him do this a few times, you realize you're in the presence of something special. It's not raw brainpower, though that helps. It's not emotional intelligence, either, though that, too, is often involved. It's deep smarts. Deep smarts are not philosophical--they're not"wisdom" in that sense, but they're as close to wisdom as business gets. You see them in the manager who understands when and how to move into a new international market, in the executive who knows just what kind of talk to give when her organization is in crisis, in the technician who can track a product failure back to an interaction between independently produced elements. These are people whose knowledge would be hard to purchase on the open market. Their insight is based on know-how more than on know-what; it comprises a system view as well as expertise in individual areas. Because deep smarts are experienced based and often context specific, they can't be produced overnight or readily imported into an organization. It takes years for an individual to develop them--and no time at all for an organization to lose them when a valued veteran walks out the door. They can be taught, however, with the right techniques. Drawing on their forthcoming book Deep Smarts, Dorothy Leonard and Walter Swap say the best way to transfer such expertise to novices--and, on a larger scale, to make individual knowledge institutional--isn't through PowerPoint slides, a Web site of best practices, online training, project reports, or lectures. Rather, the sage needs to teach the neophyte individually how to draw wisdom from experience. Companies have to be willing to dedicate time and effort to such extensive training, but the investment more than pays for itself.
Recent advances in information technology have made it possible to solve increasingly complex problems, and also to collect and store huge amounts of information. These vast quantities of data further have to be transformed into relevant value-added and "decision-quality" knowledge. It is against this background that the KDD (Knowledge Discovery in Databases), a multidisciplinary field using computer learning, artificial intelligence, statistics, database technology, expert systems, and data visualization, appeared in the early 90's. In order to assess these technologies in the medical field, we have tested some of these techniques on a large database at our disposal, named DiabCare stemming from the WHO - DiabCare program for the application of the Saint-Vincent Declaration. It contains evaluation data on the health care of patients with diabetes, and in particular, its complications. So far, data analysis has been done using classical statistical methods, and we now intend to make use of such data-mining tools as Associations Rules and Decision and Classification Trees for further exploration of this database. The results presented here show that data mining techniques can be used successfully to extract knowledge from medical databases. The results obtained using Association Rules and especially Decision Trees are very promising.
The improvement of quality in clinical care can greatly benefit form continuous benchmarking with information based on laboratory data. To be most effective, the information is best provided in the form of real-time disease management decision support. Such an approach requires advances in the implementation of hospital information systems. This article describes ground-breaking work at LDS Hospital in Salt Lake City, Utah that provides examples of major well-documented impacts of the "intelligent laboratory report" on patient outcomes.
The Integrated Academic Information Management System (IAIMS) concept is about sharing resources and information, and about improving the decision-making ability of health care professionals by integrating information. At Columbia-Presbyterian Medical Center, the IAIMS project has established an information architecture based on common, shared computing and networking resources. The institutional computing culture has been changed with increased sharing of information and, consequently, improved quality of information. Several classes of information in the areas of clinical, scholarly, administrative, basic research, and core resources have been identified for better understanding of information responsibility. Technical problems such as heterogeneity on workstation platforms and lack of universal syntactic and semantic standards for health care information exchange still impede inter-institutional sharing of information.
OBJECTIVE: To develop and internally validate a prototype multimodal artificial intelligence system for automated CEAP (Clinical, Etiological, Anatomical and Pathophysiological) classification of venous duplex ultrasound (VDUS) reports, integrating natural language processing of free-text components with computer vision analysis of hand-drawn anatomical diagrams. METHODS: Single centre retrospective observational study using routinely collected clinical data. One thousand consecutive venous duplex ultrasound reports from Cambridge University Hospitals NHS Foundation Trust, UK (July 2024 - May 2025) were labelled according to the CEAP classification, excluding the Etiological component, which could not be reliably determined from duplex reports alone. Transfer learning was applied using ClinicalBERT for text and MobileNetV3 for diagrammatic data. Clinical classes were predicted from request line text. Text- and image-based pathophysiological models were developed for four anatomical territories (Great Saphenous Vein, Small Saphenous Vein, Deep system, Perforators), combined using late fusion with probability averaging. RESULTS: The clinical CEAP model achieved accuracy of 0.91, macro-F1 of 0.82, and macro-AUC of 0.98. Pathophysiological prediction varied, with text models broadly outperforming image models. Fusion yielded heterogeneous benefits, improving SSV performance but reducing Deep system accuracy. The performance of the final pathophysiological CEAP fusion models varied across anatomical territories: accuracy ranged from 0.70-0.92 and macro-AUC from 0.80-0.92. CONCLUSION: This study demonstrates the feasibility of automated CEAP classification from VDUS reports. Despite class imbalance affecting minority class predictions, the strong discriminatory performance validates this multimodal ML model for extracting clinically meaningful information from real-world data. This approach offers potential, pending external validation, to streamline vascular services through automated triage and guideline-compliant decision making.
Computer algorithms can only produce seemingly random or pseudorandom numbers whereas certain natural phenomena, such as the decay of radioactive particles, can be utilized to produce truly random numbers. In this study, the ability of humans to generate random numbers was tested in healthy adults. Subjects were simply asked to generate and dictate random numbers. Generated numbers were tested for uniformity, independence and information density. The results suggest that humans can generate random numbers that are uniformly distributed, independent of one another and unpredictable. If humans can generate sequences of random numbers then neural networks or forms of artificial intelligence, which are purported to function in ways essentially the same as the human brain, should also be able to generate sequences of random numbers. Elucidating the precise mechanism by which humans generate random number sequences and the underlying neural substrates may have implications in the cognitive science of decision-making. It is possible that humans use their random-generating neural machinery to make difficult decisions in which all expected outcomes are similar. It is also possible that certain people, perhaps those with neurological or psychiatric impairments, are less able or unable to generate random numbers. If the random-generating neural machinery is employed in decision making its impairment would have profound implications in matters of agency and free will.
Data preprocessing is needed when real-life clinical databases are used as the data sources to learn the probabilities for dynamic decision models. Data preprocessing is challenging as it involves extensive manual effort and time in developing the data operation scripts. This paper presents a framework to facilitate automated and interactive generation of the problem-specific data preprocessing scripts. The framework has three major components: 1) A model parser that parses the decision model definition, 2) A graphical user interface that facilitates the interaction between the user and the system, and 3) A script generator that automatically generates the specific database scripts for the data preprocessing. We have implemented a prototype system of the framework and evaluated its effectiveness via a case study in the clinical domain. Preliminary results demonstrate the practical promise of the framework.
Two factors are driving a new wave of medical products. The first is the use of technology to make products "intelligent"--that is, build them not only to measure a particular parameter, like blood glucose, but to help patients and caregivers manage conditions. This allows the users of these products focus less on the technical aspects of treating a condition (e.g. calculating the proper amount of insulin to treat a given level of blood glucose) and more on the overall management of the disease. The second development is the rapid movement of devices from the doctor's office to the home. Chain drug stores carry dozens of medical devices for home use by consumers. The challenge for manufacturers and designers is to present the medical device's intelligence in a way that is palatable to the consumer. One important theme is that medical product consumers are also consumers of everything else: home electronics, appliances, clothing, etc. These consumers are applying the same decision-making processes they use when buying a blender to the process of buying a medical device. It is therefore necessary for medical product manufacturers to create devices that interact with consumers in consumer-friendly ways. Putting intelligence into a product is one thing; helping the consumer utilize and appreciate it is quite another. This chapter covers some principles to keep in mind, and discusses a framework for better design of intelligent medical products that connect with consumers on emotional and functional levels beyond simple medical efficacy.
The 68th Benzon Foundation Symposium brought together leading experts to explore the integration of mass spectrometry-based proteomics and artificial intelligence to revolutionize personalized medicine. This report highlights key discussions on recent technological advances in mass spectrometry-based proteomics, including improvements in sensitivity, throughput, and data analysis. Particular emphasis was placed on plasma proteomics and its potential for biomarker discovery across various diseases. The symposium addressed critical challenges in translating proteomic discoveries to clinical practice, including standardization, regulatory considerations, and the need for robust "business cases" to motivate adoption. Promising applications were presented in areas such as cancer diagnostics, neurodegenerative diseases, and cardiovascular health. The integration of proteomics with other omics technologies and imaging methods was explored, showcasing the power of multimodal approaches in understanding complex biological systems. Artificial intelligence emerged as a crucial tool for the acquisition of large-scale proteomic datasets, extracting meaningful insights, and enhancing clinical decision-making. By fostering dialog between academic researchers, industry leaders in proteomics technology, and clinicians, the symposium illuminated potential pathways for proteomics to transform personalized medicine, advancing the cause of more precise diagnostics and targeted therapies.
Personalized neoantigen cancer vaccine is a promising strategy for precision immunotherapy by targeting patient-specific and mutation-derived tumor antigens. Early clinical studies have demonstrated the feasibility, safety, and immunogenicity of these vaccines across multiple solid tumors, with encouraging outcomes particularly when combined with immune checkpoint blockade. However, broader clinical translation remains limited by sequential bottlenecks across the vaccine development pipeline, including false-positive neoantigen selection,  imperfect modeling of antigen processing and HLA presentation, limited prediction of T-cell receptor recognition, and challenges in formulation, delivery, and manufacturing. Artificial intelligence and advanced computational workflows are increasingly integrated into this pipeline to improve candidate prioritization and support more reproducible decision-making. In this review, we summarize clinical progress and key translational barriers in personalized neoantigen vaccination, and discuss how AI-enabled approaches may contribute across four major stages: multi-omics integration for neoantigen discovery, processing-aware HLA presentation prediction, structure-aware and TCR-informed immunogenicity modeling, and data-driven formulation optimization, particularly for lipid nanoparticle-based delivery systems. These approaches are able to help narrow biological and chemical search spaces, improve prioritization, and provide mechanistic insights into antigen presentation and immune recognition rather than replacing experimental validation. This articlefurther addresses future implementation challenges, including dataset diversity, model interpretability, prospective benchmarking, manufacturing traceability, and evolving regulatory frameworks for individualized mRNA cancer immunotherapies. Integrating computational innovation with rigorous immunological validation, scalable manufacturing, and regulatory oversight will be essential for advancing personalized neoantigen vaccines toward broader clinical implementation.
There are numerous systems and computerized procedures which have been developed and tested as a guide to clinical decisions. The methods used in the development of these systems can be divided into two large groups: mathematical and logical. None of these can be considered as better than the other in absolute, but the efficiency of each system depends on that problem is to be described and solved. The mathematical methods seem more suitable in the description of biological systems and for the selection of groups which have discriminant and sufficiently defined characteristics. Meanwhile the logical methods are better in the description and the formalization of more complicated clinical problems, characterized by uncertainty and availability of more or less numerous alternatives. It is foreseen that in the next years the systems for aided decision making will be programmed making use of methods belonging to both categories, and particularly, the expert systems will be planned using both artificial intelligence techniques and mathematical and statistical methods. The increasing frequency in the use of such systems in clinical practice will cause a reevaluation and a checking of most aspects of medical proceedings, as seen by the new methodologies or the traditional methods whose use could be driven by computer.
The practice of the decision making at the bed side especially highlights the place to be devoted to medical informatics both at the pre- and post-graduate levels. Still in a relatively recent past, say the 50s-60s, most of the medical educational efforts were delivered when watching and then imitating the medical behaviour of an older physician. The medical educators were aware that besides the formal lessons related to selected chapters of medical textbooks, there were an obvious need for better training in the ability to make sound clinical judgements. If this ability has been considered only as an artful and intuitive process neither subjected to theoretical analysis nor to be captured in a formal quantitative model, now things have changed to such an extent that it becomes broadly shared that a science of medical decision making can be reasonably founded and this threefold: 1) Upon a formulated logic, 2) The probability theory, and 3) A value theory. The first gives the hand to artificial intelligence (AI) technics, the third to medical information data bases dealing either with patients (like in hospital information systems) or with literature like MEDLINE or electronic "cookbooks". Basically the probabilistic theory is based here upon a priori probabilities related to patients informations and data and opens the way to bayesian decision making. After this little summary it is stressed that educational informatics in medicine would appear either very central or very marginal, if not optional.
Criteria for determining candidacy for an augmentative communication system are reviewed. Included is a consideration of cognitive, oral reflex, language, motor, intelligibility, emotional, chromological age, previous therapy, imitative, and environmental factors. The multiple factors are arranged on a branching type decision matrix, which yields a decision to either elect, reject, or delay implementation of an augmentative communication system. Case data demonstrating application of the decision matrix are presented.
The editor in chief of the Journal of the Canadian Dental Association explains that professional journalism alone cannot be expected to protect against questionable practice but it plays a vital role in the network of resources that define standards of excellent and ethical care. According to some, the explosion of biomedical information has created a crisis as emerging science overwhelms practitioners' capacity to intelligently evaluate it and incorporate it into practice. Research in medicine shows that journals play only a part in decisions regarding practice patterns. Some initiatives taken by the Canadian Dental Association to publish professional literature relevant to practitioners' needs to remain current in order to prevent questionable dental practice are described.
1. Patients value most the information provided to them by doctors. Doctors do not seem to be able to get messages across as satisfactorily to lower social groups as to higher social groups, since they give less time per consultation and fewer explanations to working-class patients; they are the ones that are least influenced by health advice [24] and would benefit most from being given more time and attention in consultations. 2. The community pharmacists have made great play on becoming a major information source for the public but have not yet succeeded in this role. 3. In today's society individuals will have to take greater responsibility for their own healthcare. Patients must therefore actively demand more information from their doctors and pharmacists in order to make decisions. 4. Doctors and pharmacists must recognise that patients are often unaware of what they need to know. Patients' intelligence must never be under-estimated or their knowledge and understanding over-estimated. 5. The manufacturer has a major role in providing information by means of the patient information leaflet which must be in terms that can be understood and influence the patient's health behaviour. Written information is the only information to which a patient can repeatedly refer to reinforce messages given by doctors or pharmacists. Written information is however of most value to social groups A and B and is less appreciated by other groups.
The prostate-specific antigen test has been a major factor in increasing awareness and better patient management of prostate cancer (PCA), but its lack of specificity limits its use in diagnosis and makes for poor early detection of PCA. The objective of our studies is to identify better biomarkers for early detection of PCA using protein profiling technologies that can simultaneously resolve and analyze multiple proteins. Evaluating multiple proteins will be essential to establishing signature proteomic patterns that distinguish cancer from noncancer as well as identify all genetic subtypes of the cancer and their biological activity. In this study, we used a protein biochip surface enhanced laser desorption/ionization mass spectrometry approach coupled with an artificial intelligence learning algorithm to differentiate PCA from noncancer cohorts. Surface enhanced laser desorption/ionization mass spectrometry protein profiles of serum from 167 PCA patients, 77 patients with benign prostate hyperplasia, and 82 age-matched unaffected healthy men were used to train and develop a decision tree classification algorithm that used a nine-protein mass pattern that correctly classified 96% of the samples. A blinded test set, separated from the training set by a stratified random sampling before the analysis, was used to determine the sensitivity and specificity of the classification system. A sensitivity of 83%, a specificity of 97%, and a positive predictive value of 96% for the study population and 91% for the general population were obtained when comparing the PCA versus noncancer (benign prostate hyperplasia/healthy men) groups. This high-throughput proteomic classification system will provide a highly accurate and innovative approach for the early detection/diagnosis of PCA.
Interest in decision-support programs for clinical medicine soared in the 1970s. Since that time, workers in medical informatics have been particularly attracted to rule-based systems as a means of providing clinical decision support. Although developers have built many successful applications using production rules, they also have discovered that creation and maintenance of large rule bases is quite problematic. In the 1980s, several groups of investigators began to explore alternative programming abstractions that can be used to build decision-support systems. As a result, the notions of "generic tasks" and of reusable problem-solving methods became extremely influential. By the 1990s, academic centers were experimenting with architectures for intelligent systems based on two classes of reusable components: (1) problem-solving methods--domain-independent algorithms for automating stereotypical tasks--and (2) domain ontologies that captured the essential concepts (and relationships among those concepts) in particular application areas. This paper highlights how developers can construct large, maintainable decision-support systems using these kinds of building blocks. The creation of domain ontologies and problem-solving methods is the fundamental end product of basic research in medical informatics. Consequently, these concepts need more attention by our scientific community.