Systemy Sztucznej Inteligencji
|
|---|
|
|
Przewodnik po Informatyce Medycznej, |
Systemy Sztucznej Inteligencji
|
|---|
|
|
1 Systemy interwencyjne (acute care), do przypadków nagłychWyatt J. "The evaluation of clinical decision aids: a discussion of methodology used in the ACORN project", Lecture Notes in Medical Informatics 1987; 33: 15- 24
Planning of an adequate nutritional support for maintaining the metabolic needs of sick newborn infants is time consuming, needs experts' knowledge and involves the risk of introducing possibly fatal errors. Recent systems used for composing parenteral nutrition solutions mainly support the calculation and the documentation process and cannot easily be adapted for neonates. Computerized expert system technology may help to develop time saving solutions to a given problem and to avoid errors within certain limits. We therefore developed an interactive expert system for calculating the composition of parenteral nutrition solutions (PNS) for newborn infants.
The knowledge base of the expert system consists of the rules for composing the PNS according to heuristic rules used at the cooperating NICU. Applying these rules, the daily fluids, electrolytes, vitamins, and nutritional requirements were calculated according to the estimated needs, the patient's body weight, the age, and the individual tolerance. The requirements were also corrected according to the daily measurements of serum electrolytes, triglycerides and protein if available. Glucose supply was adjusted depending on the type of venous access used (peripheral or central venous line), on the glucose tolerance and on the total fluid allowances. Finally, the PNS was reduced according to the proportion of oral feedings. The program works interactively asking for relevant data, calculating the PNS, and displaying the results. The physician has the choice to adjust calculated values according to special clinical requirements. The final output is a PNS schedule that can be used directly in the case history of neonates. Possible input and dosage errors are eliminated by methods of data validation using body weight and age dependent thresholds.
A knowledge acquisition module supports updating of thresholds, input of medication of new bypass and new oral feeding products. VIE-PNN was developed on an IBM compatible PC. Currently, a practical clinical evaluation of VIE-PNN is performed at the ICU.
The project is a joint cooperation of the Austrian Research Institute for Artificial Intelligence (OFAI), the Department of Medical Cybernetics and Artificial Intelligence (IMKAI), and the Neonatal Intensive Care Unit (NICU) of the Department of Pedriatics of the University of Vienna:
1.Dojat M. and Pachet F., An extendable knowledge-based system for the control of mechanical ventilation, in Proc. 14th IEEE-EMBS, Paris, pp. 920-921, 1992.
2.Dojat M. and Pachet F., Representation of a medical expertise using the Smalltalk environment: putting a prototype to work, in TOOLS 7, G. Heeg, B. Magnusson and B. Meyer, Ed., New York: Prentice Hall, pp. 379-389, 1992.
3.Dojat M., Brochard L., Lemaire F. and Harf A., A knowledge-based system for assisted ventilation of patients in intensive care, International Journal of Clinical Monitoring and Computing, vol. 9, pp. 239-250, 1992.
4.Dojat, M. and Sayettat, C. Aggregation and forgetting: two key mechanisms for across-time reasoning in patient monitoring. In ""Proceedings of AAAI spring symposium. Artificial Intelligence in Medicine: Interpreting Clinical Data", (I. Kohane and S. Uckun, Eds.), pp. 27-31. AAAI Technical Report SS-94-01, Stanford University (Ca), 1994.
5.Dojat M. and Sayettat C., A realistic model for temporal reasoning in real-time patient monitoring, Applied Artificial Intelligence, vol. 10, n°2, 1996, (to appear).
6.Dojat M., Harf A., Touchard D., Laforest M., Lemaire F. and Brochard L., Evaluation of a knowledge-based system providing ventilatory management and decision for extubation, American Journal of Respiratory and Critical Care Medicine, 1996 (to appear).
The VentEx system has been built to support ventilator therapy management using knowledge-based system technology. Development started with an early prototype system called KUSIVAR [I] which dealt with knowledge representation and knowledge acquisition research using the Knowledge Engineering Environment (KEE). A domain specific tool called KAVE [II] was later developed to facilitate the knowledge acquisition process. Then a PC-based on-line system (VentEx) was built [III] as an integrated knowledge-based system in the clinical environment using Nexpert Object. Results of the evaluation work [IV] indicate the usefulness of KAVE, and there was a high consensus between the doctors and VentEx according to a "gold" standard [V].
[I] Shahsavar N, Frostell C, Gill H, Ludwigs U, Matell G and Wigertz O. Knowledge Base Design for Decision Support in Respirator Therapy. International Journal of Clinical Monitoring and Computing, 1989, 6:223-231.
[II] Shahsavar N, Gill H, Wigertz O, Frostell C, Matell G and Ludwigs U. KAVE: A Tool for Knowledge Acquisition to Support Artificial Ventilation. International Journal of Computer Methods and Programs in Biomedicine, 1991, 34: 115-123.
[III] Shahsavar N, Gill H, Ludwigs U, Carstensen A, Larsson H, Wigertz O and Matell G. VentEx: An On-Line Knowledge-Based System to Support Ventilator Management. Technology and Health Care, 1994, 1:233-243.
[IV] Shahsavar N, Ludwigs U, Blomqvist H, Gill H, Wigertz O and Matell G. Evaluation of a Knowledge-Based Decision-Support System for Ventilator Therapy Management. Artificial Intelligence in medicine, 1995, 7:37-52.
[V] Nosrat Shahsavar. Design, Implementation and Evaluation of a Knowledge-Based System to Support Ventilator Therapy Management. Linkoping Studies in Science and Technology, PhD thesis 317, Department of Medical Informatics, Linkoping University, Sweden, 1993.
SJ. DARMONI, P. MASSARI, JM. DROY, E. MOIROT, J. LE ROY. Functional evaluation of SETH: an expert system in clinical toxicology Proceedings of the 5th Conference on Artificial Intelligence in Medicine Europe, P. Barahona, M. Stefanelli, J. Wyatt (Eds), pp 231-238 (Pavie, Italie, Juin 1995).
SJ. DARMONI, P. MASSARI, JM. DROY, T. BLANC, F. MORITZ, N. MAHE, J. LEROY. From general reasoning in drug poisoning to specific attitudes in human and in SETH. Computer as an aid in poison centres, Lille, Décembre 1995.
P. MASSARI, SJ. DARMONI, JM. DROY, T. BLANC, F. MORITZ, N. MAHE, J. LEROY. Seth, an expert system in drug poisoning: five years later. Computer as an aid in poison centres, Lille, Décembre 1995.
SJ. DARMONI, P. MASSARI, JM. DROY, E. MOIROT, J. LE ROY. SETH: an expert system for the management on acute drug poisoning in adults. Comput. Methods Programs Biomed. 1993; 43: 171-176.
2 Laboratory Systems
Miejsce stosowania:
Information about the St. Bartholomew's site can also be obtained from Dr. John Amess, Consultant Haematologist, Department of Haematology, St. Bartholomew's Hospital, West Smithfield, London, EC1A 7BE, UK, Telephone: 44 171 601-8204, Fax: 44 171 601-8200.
Information regarding the IZASA-Coulter CITOTECA Workstation and the sites in Spain and Portugal can be obtained from Dr. Ramon Simon, Haematology Division, IZASA, S.A., Aragon 90, 08015 Barcelona, SPAIN, Telephone: 34 3 4010101, Fax: 34 3 4010230.
The Kontakt at PGP (for the site in Belgium) is Renato PROTTI, Product Specialist, Rue Driesstraat 175, Bruxelles B-1200, Belgium, Telephone: 32 2 770 62 22, Fax: 32 2 770 92 25, E-mail: 100653.2230@compuserve.com .
(1) At St. Bartholomew's, FACULTY is running on a five computer network with a bi-directional interface to the laboratory's LIS using a custom Coulter haematology communications server (HCS). Specimen orders are passed from the LIS to the HCS. The HCS filters the haematology specimens as they are run on a STKS. Normal specimens are routed directly to the LIS. Abnormal specimens artion, to the PGP system which downloads STKS data to the LIS. Abnormal specimens are stored in Professor Petrushka's database. Peripheral blood film review is performed using Professor Petrushka and the results are passed back to the LIS. Flow cytometry immunophenotyping results are stored in Professor Fidelio's database. We are currently setting up Professor Belmonte to process bone marrow reports;
(2) At the three sites in Spain and Portugal, FACULTY is installed on the IZASA-Coulter CITOTECA (R) Workstation which includes a mini-LIS (Modulab Plus), instrument interfaces, and a facility for capturing images from a video camera/microscope attached to the workstation. The FACULTY Paradox databases are interfaced to the dBase tables maintained by Modulab Plus;
(3) At Cliniques Universitares Mont-Godinne, FACULTY is interfaced, via a network connection, to the PGP system which downloads STKS data to the LIS.
(1) Diamond LW, Nguyen DT: Expert systems in laboratory haematology. In: Lewis SM, Koepke JA (eds), Haematology Laboratory Management and Practice. Butterworth-Heinemann, Oxford, 1995, pp.43-51.
(2) Diamond LW, Nguyen DT, Andreeff M, Maiese RL, Braylan RC: A knowledge-based system for the interpretation of flow cytometry data in leukemias and lymphomas. Cytometry 17:266-373, 1994.
(3) Diamond LW, Mishka VG, Seal AH, Nguyen DT: Multiparameter interpretive reporting in diagnostic laboratory hematology. International Journal of Biomedical Computing 37:211-24, 1994.
(4) Nguyen DT, Diamond LW, Priolet G, Sultan C: Expert system design in hematology diagnosis. Methods of Information in Medicine 31:82-9, 1992.
Renal function varies over time and can be estimated as a function of calculated creatinine clearance. DoseChecker is an expert system which monitors patients with active orders for drugs known to require careful dosing. Using parameters such as patient weight and serum creatinine, DoseChecker calculates creatinine clearance and applies a set of dosing guidelines developed by pharmacokinetic experts to determine if the dosing is appropriate. If it does not fall within established guidelines, an alert is generated for a pharmacist, who then consults with the patient's attending physician to determine whether the dosage should be adjusted.
DoseChecker uses a relational database containing patient demographic information and clinical data such as serum creatinine measurements and drug orders. Suspected dosing violations are stored so that trends can be detected.
1. Kahn MG, Steib SA, Fraser VJ, Dunagan WC. An expert system for culture- based infection control surveillance. In: Safran C, ed. Proceedings Symposium on Computer Applications in Medical Care. New York, NY: McGraw Hill, 1993:171-5.
2. Kahn MG, Steib SA, Spitznagel EL, Dunagan WC, Fraser VJ. Improvement in User Performance Following Development and Routine Use of an Expert System. In: Greenes RA, Peterson HE, Protti DJ, eds. MEDINFO '95. Edmonton Alberta, Canada: International Medical Informatics Association / Healthcare Computing & Communications Canada, Inc., 1995:1064-67.
KP Adlassnig, W Horak, Routinely-used, automated interpretive analysis of hepatitis A and B serology findings by a medical expert system, Proc. Medical Informatic Europe `90, R. O'Moore et al. (eds), Lecture Notes in Medical Informatics, 40, Springer-Verlag, 313-318.
Edwards G, Compton P, Malor R, Srinivasan A, Lazarus L. PEIRS: a pathologist maintained expert system for the interpretation of chemical pathology reports. Pathology 1993;25:27-34
Kunz, J.C., R.J. Fallat, D.H. McClung, et. al., "Automated interpretation of pulmonary function test results". Proceedings of Computers in Critical Care and Pulmonary Medicine, IEEE Press, 1979.
Morrell RM, Wasilauskas BL, Winslow RM. Personal computer-based expert system for quality assurance of antimicrobial therapy. Am J Hosp Pharm. 1993;50:2067-73.
C Trendelenburg, B. Pohl, Pro. M. D.: Medical Diagnostics with Expert Systems An introduction with diskettes to the expert system shell Pro. M. D. Publisher: MEDISOFT 1995, 4. edition 180 pages with 3,5' diskette ISBN: 3-931296-04-0 145,-- DM
December 94 issue of the Journal of laboratory medicine (LaboratoriumsMedizin) contains 6 papers on Pro.M.D.:
3 Educational Systems
4 Quality Assurance and Administration
P.J. Byrns, D.C. Lezotte, and J. Bondy, "Influencing the cost- effectiveness of prescribing using claims-based information: a randomized trial", submitted to J. Am. Med. Assoc.
We have developed an expert system called Reportable Diseases, which applies state Public Health Department culture-based criteria for detecting "significant" infections, which are required to be reported to the state. Reportable Diseases has been deployed at Barnes and Jewish Hospitals, tertiary-care teaching hospitals, since February 1995.
Microbiology culture data from the hospital's laboratory system are monitored by Reportable Diseases. Using a rulebase consisting of criteria developed by the state Public Health Department, Reportable Diseases scans the culture data and generates an "alert" to the Infection Control staff when a culture representing a "reportable" infection is detected.
George Hripcsak, Peter Ludemann, T. Allan Pryor, Ove B. Wigertz, Paul D. Clayton. Rationale for the Arden Syntax. Computers and Biomedical Research 1994;27:291-324.
T. Allan Pryor, George Hripcsak. Sharing MLM's: an experiment between Columbia-Presbyterian and LDS Hospital. In: Safran C, editor. Proceedings of the Seventeenth Annual Symposium on Computer Applications in Medical Care; 1993 Oct 30-Nov 3; Washington, D. C. New York: McGraw-Hill, Inc., 1994; 399-403.
George Hripcsak. Monitoring the Monitor: Automated Statistical Tracking of a Clinical Event Monitor. Computers and Biomedical Research 1993;26:449-66.
George Hripcsak, James J. Cimino, Stephen B. Johnson, Paul D. Clayton. The Columbia-Presbyterian Medical Center decision-support system as a model for implementing the Arden Syntax. In: Clayton PD, editor. Proceedings of the Fifteenth Annual Symposium on Computer Applications in Medical Care; 1991 Nov 17-20; Washington, D.C. New York: McGraw-Hill, Inc., 1992; 248-52.
G. Hripcsak, P.D. Clayton, J.J. Cimino, S.B. Johnson, C. Friedman. Medical decision support at Columbia-Presbyterian Mecial Center. In: Timmers T, Blum BI, editors. Software Engineering in Medical Informatics. Amsterdam: North-Holland, 1991, pp. 471-9.
5 Medical Imaging
6 Decison Support Systems
Around 300 questions could be asked of the patient; however, the system guides the nurse to ask 20 to 40 questions relevant to a particular patient. The progress note, organized in the SOAP format, is reviewed by the physician with the patient. The physician could also review the clinical data, weigh the suggestions from the system, and modify the Assessment or Plan sections. The Subjective and Objective sections could also be modified but rarely needed to be. Without the system a physician spent 21.35 min (+/- 0.95 sem, N=140) with the patient. With the system, the nurse spent 14.95 min (+/- 0.81 sem, N=27), and the physician spent 7.4 min (+/- 0.68 sem, N=27). Physician time was cut by about 66%. Using 1994 VA salaries for nurses and physicians, we have shown that the system reduced cost by about 40%.
We have compared the quality of the progress note generated by physicians to the computer generated note. Using a scoring system that divides the note data into essential and bonus categories, we found that the computer note quality was higher (95.5, +/-8.19 sd, N=12) compared to a physician's hand written note ( 85.2, +/-9.11 sd, N=24; p < 0.01).
Our informal assessment of the system is that it was well accepted by our physicians, nurses, and patients. Our physicians were willing to give up time on routine cases in exchange for more time on more difficult cases. Nurses liked the system because they could work at a higher level of expertise and spend more time with the patient. Patients seemed willing to accept the system even though they were waiting for two interviews (nurse and physician).
The system uses an object-oriented architecture and is divided into modules which contain both rules and data, and communicate with each other by passing conclusions. We organized the objects by physiological system. The system runs on a PC under Windows 3.11 and was constructed using ToolBook (Asymetric) for the user interface, Nexpert Object (Neuron Data) for the inferencing engine, and DBase III (Borland) for data storage. The system contained about 25 screens, 250 rules, and 300 data fields in about 30 files.
Doller, H. J., Hostetler, W.E., Krishnamurthy, K., and Peterson, L.L., Epileptologists' Assistant: A Cost Effective Expert System, SCAMC 17:384-388, 1994.
Doller, H. J., Hostetler, W. E., and Peterson, L. L.: Expert Systems Decrease the Cost While Increasing the Quality of Out Patient Clinical Encounters AMIA 1995 Spring Congress, Cambridge, MA, June 24-28, 1995.
Doller, H. J., Hostetler, W., Krishnamurthy, K., and Peterson, L.L.: Expert Systems: Cost Effective Patient Data Gathering Tools for the Electronic Medical Record. AAAI Spring Symposium, St Louis, May 9-15, 1993.
Hostetler, W.E. and Doller, H. J.: Epileptologists' Assistant: an Expert System for Epilepsy Clinic Improves Progress Note Quality While Decreasing Visit Cost, Epilepsia 35:(supp. 8) 45, 1994.
Hostetler, W., Doller, H. J., Krishnamurthy, K., and Peterson, L.L.: Epileptologist's Assistant: A Cost Effective Expert System for Clinical Medicine. First World Conference on Computational Medicine, Public Health and Biotechnology, Austin, Texas., April 24-26, 1994.
Hostetler, W., Krishnamurthy, K., Peterson, L.L., and Doller, H. J., The Physician's Interface to Epileptologist's Assistant - A Cost Effective Expert System, SCAMC 17:944, 1994.
We have built an expert system and a knowledge acqusition component which is routinely applied to more than 200 prototypical Opiss of dysmorphic syndromes. Prototypes consist of simple feature lists. The catalogue of features has 823 entries. The patient data management component of the system supports the handling of all clinical data.
We evaluated our approach using 903 patients and 229 different prototypes of dysmorphic syndrome which have been collected over many years in a pediatric clinic at the University of Munich. As a result we observed good sensitivity for the system, comparable decisions to the involved physicians and more precise and enhanced knowledge on dysmorphic syndromes. One of the major advantages of case-based systems is that the semi-automatically and incrementally generated prototypes are highly site-specific i.e. are adapted to the set of diseases specific for the patients seen in this pediatric clinic.
Up to now the system has been used on about 3000 patients. All knowledge about these patients is integrated into the knowledge-base. Up to three physicians have been used MDDB since 1988 daily.
The program was designed and written through joint cooperation between the Department of Engineering Mathematics and the Department of Child Dental Health, University of Bristol. Development was funded by an MRC Grant.
Whilst the mechanical side of treatment is relatively straightforward, success depends upon adopting an appropriate treatment plan. Studies have shown that less that half the treatment plans adopted by practitioners are ideal and this considerably compromises the standard of result which is obtained. Jeremiah has been shown to improve on the ability of practitioners to select cases for suitable for treatment with removable orthodontic appliances and to identify those requiring referral for more specialised treatment.
Mackin N, Stephens CD, (1997). Development and testing of a fuzzy expert system - an example in orthodontics in proceedings of fuzzy logic: applications and future directions, pp61-71. Unicom Seminars Ltd, Uxbridge, Middlesex.
Richmond S, Shaw WC, Stephens CD, O'Brien KD, Brooke PH, Roberts C, Andrews M, (1993) Orthodontics in the General Dental Service of England and Wales: a critical assessment of standards. British Dental Journal, 174: 315-329.
Sims-Williams JH, Brown ID, Matthewman A, Stephens CD, (1987) A computer controlled expert system for orthodontic advice. British Dental Journal, 163: 161-169.
Sims-Williams JH, Mackin N, Stephens CD, (1994) Lessons learnt from the development of an orthodontic expert system in Neural networks in medicine and healthcare. Ifeachor CD, Rosen KG (eds), pp410-414, University of Plymouth.
Stephens CD, Drage KD, Richmond S, Shaw WC, Roberts CT, Andrews M, (1993). Consultant opinion on orthodontic treatment plans devised by dental practitioners: a pilot study. Journal of Dentistry, 21: 355-359.
Stephens CD Mackin N, Sims-Williams JH, (1996) The development and validation of an orthodontic expert system. British Journal of Orthodontics, 23: 1-9.
Orthoplanner was developed by cooperation between the Department of Engineering Mathmatics and Department of Child Dental Health, University of Bristol and Team Management Systems, Aylesbury, Buckinghamshire, with support from 2 SMART Awards (Small Firms Merit Award for Research and Technology).
Orthoplanner is a Windows based program. It uses a number of techniques including rulebase reading, but forward and backward chaining and fuzzy logic based representations of orthodontic knowledge (Mackin 1992). Extensive use is made of interactive graphics to input clinical data. In addition to treatment planning advice, the program provides extensive clinical support including instructions to patients, pre-formed letters and a 200 page hypertext manual with 1000 supporting Literatura.
Mackin N, (1992). The development of an expert system for planning orthodontic treatment. PhD Thesis, University of Bristol.
Stephens CD, Mackin N, (1998). The validation of an orthodontic expert system rulebase for fixed appliance treatment planning. European Journal of Orthodontics (accepted for publication).