Biomedical subjects

C E Kahn

Publications and source records attributed to C E Kahn.

At least 37 records · Page 2Linked to original sources

PubMed1995

CHORUS: a computer-based radiology handbook for international collaboration via the World Wide Web.

To facilitate collaboration among physicians, a computer-based radiology handbook was developed and published electronically via the World Wide Web on the Internet. This system, called CHORUS (Collaborative Hypertext of Radiology), allows physicians without computer expertise to read documents, contribute knowledge, and critically review the handbook's content by using a simple, graphical user interface from virtually any type of computer system. CHORUS contains 1,168 "note-card" documents that describe radiologic findings; differential diagnoses; technical information; and pertinent anatomy, pathology, and physiology. Documents are indexed by title and by organ system and are linked to related documents. Data entry forms allow physicians to comment on published documents, submit new documents, and review submitted documents. CHORUS uses public-domain technologies to present useful, easily accessible knowledge for education and clinical decision making, and it provides a medium for international medical collaboration via the Internet.

Computer Communication Networks↗

PubMed1995

Preliminary investigation of a Bayesian network for mammographic diagnosis of breast cancer.

Bayesian networks use the techniques of probability theory to reason under conditions of uncertainty. We investigated the use of Bayesian networks for radiological decision support. A Bayesian network for the interpretation of mammograms (MammoNet) was developed based on five patient-history features, two physical findings, and 15 mammographic features extracted by experienced radiologists. Conditional-probability data, such as sensitivity and specificity, were derived from peer-reviewed journal articles and from expert opinion. In testing with a set of 77 cases from a mammography atlas and a clinical teaching file, MammoNet performed well in distinguishing between benign and malignant lesions, and yielded a value of 0.881 (+/- 0.045) for the area under the receiver operating characteristic curve. We conclude that Bayesian networks provide a potentially useful tool for mammographic decision support.

Bayes Theorem↗

PubMed1995

Decision-theoretic refinement planning: a new method for clinical decision analysis.

Clinical decision analysis seeks to identify the optimal management strategy by modelling the uncertainty and risks entailed in the diagnosis, natural history, and treatment of a particular problem or disorder. Decision trees are the most frequently used model in clinical decision analysis, but can be tedious to construct, cumbersome to use, and computationally prohibitive, especially with large, complex decision problems. We present a new method for clinical decision analysis that combines the techniques of decision theory and artificial intelligence. Our model uses a modular representation of knowledge that simplifies model building and enables more fully automated decision making. Moreover, the model exploits problem structures to yield better computational efficiency. As an example we apply our techniques to the problem of management of acute deep venous thrombosis.

Artificial Intelligence↗

PubMed1995

Optimizing, diagnostic, and therapeutic strategies using decision-theoretic planning: principles and applications.

OBJECTIVE: Decision-theoretic planning is a new technique for selecting optimal actions. The authors sought to determine whether decision-theoretic planning could be applied to medical decision making to identify optimal strategies for diagnosis and therapy. METHODS: An existing model of acute deep venous thrombosis (DVT) of the lower extremities--in which 24 management strategies were compared--was converted into a set of conditional-probabilistic actions for use by the DRIPS decision-theoretic planning system. Actions were grouped into an abstraction/decomposition hierarchy. A utility function was defined in accordance with the existing DVT management model to incorporate the costs and risks of the diagnostic tests and treatments. RESULTS: From 18 primitive actions (such as "perform venography" and "treat if venography shows thigh DVT"), a total of 312 possible concrete plans were encoded within the abstraction/decomposition hierarchy. The DRIPS planning system used abstraction techniques to eliminate 136 possible plans (44%) from consideration. It determined that, given the parameters specified, the most cost-effective management strategy was "no tests, no treatment." This result differed from the published result of "perform ultrasonography, treat if positive." In reviewing the original article, it was determined that DRIPS had revealed an error in the manually constructed decision trees used in that manuscript. At values of $75,000 and greater for the cost of death, the optimal strategy became "impedance plethysmography (IPG), don't wait, perform venography if IPG is positive, and treat only if venography shows thigh DVT." CONCLUSION: Decision-theoretic planning is applicable to medical decision making and may be an extremely useful technique for complex decisions. The use of inheritance abstraction makes the technique computationally tractable for complex planning problems, and the modular nature of the data entry may help eliminate errors that appear in manually encoded decision trees.

Computer Simulation↗

PubMed1994

Case-based reasoning and imaging procedure selection.

RATIONALE AND OBJECTIVES: Case-based reasoning, an artificial intelligence technique for learning and reasoning from experience, has shown great potential for use in decision support systems. The authors developed and tested a prototype case-based decision support system to explore the applicability of this technique to the selection of diagnostic imaging procedures. METHODS: A case-based system, ProtoISIS, was developed based on the Protos learning apprentice. ProtoISIS learned the domain of ultrasonography and body computed tomography by reviewing 200 consecutive cases of actual requests for imaging procedures. ProtoISIS was tested by using it to classify four sets of 25 cases of actual imaging procedure requests. RESULTS: ProtoISIS correctly classified 72% of the imaging-procedure requests. Its performance improved as it gained experience: in the last two test series, it correctly classified 84% of the cases presented. CONCLUSIONS: Case-based reasoning can be applied successfully to the selection of diagnostic imaging procedures and holds potential for use in clinical decision support aids. Further work is necessary to realize a clinically useful system.

Artificial Intelligence↗

PubMed1994

A Bayesian network model for radiological diagnosis and procedure selection: work-up of suspected gallbladder disease.

Bayesian networks, a technique for reasoning under uncertainty, currently are being developed for application to medical decision making. To explore their usefulness for radiologic decision support, a Bayesian belief network was constructed in the domain of hepatobiliary disease. The network model's nodes represent diagnoses, physical findings, laboratory test results, and imaging study findings. The connections between nodes incorporate conditional probabilities, such as sensitivity and specificity, to represent probabilistic influences. Statistical data were abstracted from peer-reviewed journal articles on hepatobiliary disease, and a network was created to reflect the data. The network successfully determined the a priori probabilities of various diseases, and incorporated laboratory and imaging results to calculate the a posteriori probabilities. The most informative examination was identified, that is, the laboratory study or imaging procedure that led to the greatest diagnostic certainty. Bayesian networks represent a very promising technique for decision support in radiology: they can assist physicians in formulating diagnoses and in selecting imaging procedures.

Adult↗

PubMed1994

Artificial intelligence in radiology: decision support systems.

Computer-based systems that incorporate artificial intelligence techniques can help physicians make decisions about their patients' care. In radiology, systems have been developed to help physicians choose appropriate radiologic procedures and to formulate accurate diagnoses. These decision support systems use techniques such as rule-based reasoning, artificial neural networks, hypertext, Bayesian networks, and case-based reasoning. This article reviews these artificial intelligence techniques, describes their application in radiology, and discusses the role that decision support systems may play in radiology's future.

Artificial Intelligence↗

PubMed1994

Hepatic helical CT: contrast material injection protocol.

PURPOSE: To develop and compare contrast material injection protocols suitable for hepatic helical computed tomography (CT). MATERIALS AND METHODS: Monophasic and biphasic helical CT were performed with contrast material with an iodine load of 50 g at 3 mL/sec for 60 seconds or at 5 mL/sec for 10 seconds and 2 mL/sec for 65 seconds, respectively. In 58 men and 51 women, aged 22-77 years, aortic and hepatic enhancement curves were constructed from a cluster acquisition with a slip-ring scanner operating in a nonhelical mode. RESULTS: The monophasic protocol produced a higher peak aortic enhancement (180 HU +/- 47 [+/- 1 standard deviation]) than the biphasic protocol (150 HU +/- 24). Peak hepatic enhancement (63-64 HU +/- 15) was equivalent. Calculated equilibrium time for the monophasic protocol was 95.1 seconds and for the biphasic protocol was 101.4 seconds. The contrast enhancement index differed only marginally between the two protocols (P < .4). CONCLUSION: Monophasic and biphasic protocols produced equivalent results when tailored for the shorter temporal window of a rapid-sequence helical acquisition.

Adult↗

PubMed1994

Generating explanations and tutorial problems from Bayesian networks.

We present a system that generates explanations and tutorial problems from the probabilistic information contained in Bayesian belief networks. BANTER is a tool for high-level interaction with any Bayesian network whose nodes can be classified as hypotheses, observations, and diagnostic procedures. Users need no knowledge of Bayesian networks, only familiarity with the particular domain and an elementary understanding of probability. Users can query the knowledge base, identify optimal diagnostic procedures, and request explanations. We describe BANTER's algorithms and illustrate its application to an existing medical model.

Algorithms↗

PubMed1994

Planning diagnostic imaging work-up strategies using case-based reasoning.

ISIS is a developmental decision support system that helps physicians select diagnostic imaging procedures. It uses case-based reasoning, an artificial-intelligence approach that emphasizes reasoning and planning from prior experience. The development, training, and evaluation of a prototype system were used to guide the development of ISIS. To realize a clinically useful system, particular emphasis has been placed on increasing the depth and breadth of case-based knowledge, enhancing the explanatory capabilities of the system, and refining the human-computer interfaces to include a critiquing approach.

Algorithms↗

PubMed1993

Graphical knowledge presentation in a MUMPS-based decision-support system.

Conventional knowledge-based medical expert systems present information and elicit responses in the form of text. PHOENIX, a decision-support system designed to help non-radiologist physicians select diagnostic imaging procedures, offers a graphical user interface. The system constructs and displays algorithms, or flowcharts, from the rules in its knowledge base. Users can view the flowcharts and move through them by answering questions at branch points. The system provides detailed textual explanations of the rules and descriptions of the imaging procedures. The system is written in MUMPS, a common programming language for biomedical information systems, and is easily incorporated into clinical computer systems.

Algorithms↗

PubMed1993

Magnetization transfer imaging of the abdomen at 0.1 T: detection of hepatic neoplasms.

Magnetization transfer (MT) techniques have been proposed as a method of increasing contrast in MR images. To evaluate the feasibility of MT imaging of the abdomen at 0.1 T and to assess the clinical utility of this technique, the authors studied tissue contrast with a gradient-echo pulse sequence and an MT sequence in four normal volunteers, and in 17 patients with known primary or secondary neoplasms of the liver. The MT technique increased contrast between the liver and other tissues such as spleen, skeletal muscle and subcutaneous fat. The technique also produced increased contrast between hepatic tumors and normal liver parenchyma in gradient-echo images.

Abdomen↗

PubMed1993

Computed tomography for nontraumatic headache: current utilization and cost-effectiveness.

A retrospective study was performed at two teaching hospitals--one in the United States and one in Canada--to determine the results of computed tomography (CT) examinations of the head in patients with nontraumatic headache. Of 1111 examinations performed over a 3-year period, 120 (10.8%) demonstrated an acute intracranial abnormality, such as hemorrhage, infarction or tumour; the frequency of such abnormalities was highest among inpatients and subjects over 40 years of age. Cranial and extracranial abnormalities, such as sinusitis and metastases to the calvarium, were found in 40 (3.6%) of the cases. Chronic abnormalities, such as cerebral atrophy or remote infarction, were the most significant findings in 202 (18.2%) of the cases. The cost of finding each case of acute intracranial abnormality was $5962 (US); for subarachnoid hemorrhage among patients in the emergency department, it was $15,837 (US).

Adolescent↗

PubMed1992

Magnetization transfer contrast imaging of the human leg at 0.1 T: a preliminary study.

Magnetization transfer contrast imaging is an MR technique that capitalizes on interactions between the protons of mobile and macromolecularly bound water molecules. Studies to date, conducted primarily on 4.7 T and 1.5 T MR systems, have yielded results unique from conventional T1- and T2-weighted imaging studies. In this study, performed on a 0.1 T device, a section of lower leg was imaged in 20 normal human subjects and one patient with muscular dystrophy, using both a standard 500/22 gradient-echo sequence and a 500/22 gradient-echo sequence combined with off-resonance radio frequency irradiation designed to elicit magnetization transfer contrast. Results of the two techniques were compared. Our findings suggest that magnetization transfer contrast imaging is feasible at 0.1 T, and that this technique allows reproducible tissue characterization and improves contrast between certain tissues.

Adipose Tissue↗

PubMed1992

Computer-aided detection of diffuse liver disease in ultrasound images.

The authors are developing an automated method of determining measures of texture in liver ultrasound images to improve the accuracy of diagnosis of diffuse liver abnormalities. In each digitized image, the background trend of pixel values is estimated to isolate the underlying pattern of liver texture. After correcting for background trend, the root mean square (RMS) variation and the first moment of the power spectrum are calculated; these measures have been applied successfully to texture analysis of digital chest radiographs. Background trend-corrected measurements detected statistically significant differences in digitized ultrasound images of 11 normal and 11 abnormal livers. Without correction for background trend, the measures are unable to distinguish normal from abnormal liver texture. The authors also investigated the effect on the texture measures of varying several ultrasound imaging parameters in normal patients and in an ultrasound phantom.

Adult↗

PubMed1991

A radiology hypertext system for education and clinical decision making.

Hypertext is a computer-based means of organizing information that allows one to explore the connections among related subjects. FACT/FILE is a hypertext computer system that provides access to a wide variety of information of interest to radiologists and radiology residents. Information is presented as notecards, or frames, which can contain textual information about relevant anatomy, physiology and pathology, as well as differential diagnosis "gamut" listings. More than 800 frames are available. Users can choose frames from an alphabetical index produced automatically from key words in each frame's title, or from a hierarchically organized subject index. Links between frames allow users to explore related frames and then return to the original frame. Physicians can add new frames to the system, so that specialized knowledge can be shared. This system has been implemented as a module of the radiology information system at the University of Chicago Medical Center.

Computer-Assisted Instruction↗

PubMed1991

Vena Tech vena cava filter: experience and early follow-up.

Vena caval filters, such as the Vena Tech filter, that employ low-profile introducer systems have provided physicians with a variety of options for percutaneous placement. From April 1989 to April 1990, 81 patients underwent percutaneous placement of the Vena Tech filter at the authors' institution. Follow-up has been obtained to evaluate the filter with regard to the prevention of pulmonary embolism, the maintenance of caval patency, and mechanical stability. Two cases of pulmonary embolism have been seen following filter placement. Three cases of caval thrombosis have occurred, with recanalization of the cava seen in two of these cases. There have been one broken filter and one case of incomplete filter opening. Limited filter tilting and migration have occurred, though in no case has filter tilt or migration been clinically significant. This experience with the Vena Tech filter suggests that it is safe and effective for the prevention of pulmonary embolism.

Adolescent↗

PubMed1991

Validation, clinical trial, and evaluation of a radiology expert system.

The PHOENIX Radiology Consultant is a rule-based expert system which assists physicians in planning radiological work-up strategies. This article describes the methods used to create and validate the system's knowledge base. The feasibility and acceptability of PHOENIX were tested for two years in a clinical trial. During this period, the system was used 1,421 times, an average of 13.7 times per week, primarily by medical students and nonradiologist physicians. Much of the system's use occurred at night and on weekends, when the radiology department was not fully staffed. Several physicians were enlisted to further evaluate the utility of the system. The results of their evaluation indicate that an expert system that helps physicians select diagnostic-imaging studies can serve as a useful and informative component of a radiology information system, and is particularly useful for medical students and physicians in training.

Algorithms↗