Computer-based models of medical decision making account for a large portion of clinical computing efforts. This article reviews representative examples from each of several major medical computing paradigms. These include 1) clinical algorithms, 2) clinical databanks that include analytic functions, 3) mathematical models of physical processes, 4) pattern recognition, 5) Bayesian statistics, 6) decision analysis, and 7) symbolic reasoning or artificial intelligence. Because the techniques used in the various systems cannot be examined exhaustively, the case studies in each category are used as a basis for studying general strengths and limitations. It is noted that no one method is best for all applications. However, emphasis is given to the limitations of early work that have made artificial intelligence techniques and knowledge engineering research particularly attractive. We stress that considerable basic research in medical computing remains to be done and that powerful new approaches may lie in the melding of two or more established techniques.

Knowledge engineering for medical decision making: A review of computer-based clinical decision aids

A number of biomarkers of inflammatory and metabolic pathways are individually related to higher risk of colorectal cancer (CRC); however, the association between biomarker patterns and CRC incidence has not been previously evaluated. Our study investigates the association of biomarker patterns with CRC in a prospective nested case-control study within the European Prospective Investigation into Cancer and Nutrition (EPIC). During median follow-up time of 7.0 (3.7-9.4) years, 1,260 incident CRC cases occurred and were matched to 1,260 controls using risk-set sampling. Pre-diagnostic measurements of C-peptide, glycated hemoglobin, triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), C-reactive protein (CRP), reactive oxygen metabolites (ROM), insulin-like growth factor 1, adiponectin, leptin and soluble leptin receptor (sOB-R) were used to derive biomarker patterns from principal component analysis (PCA). The relation with CRC incidence was assessed using conditional logistic regression models. We identified four biomarker patterns 'HDL-C/Adiponectin fractions', 'ROM/CRP', 'TG/C-peptide' and 'leptin/sOB-R' to explain 60 % of the overall biomarker variance. In multivariable-adjusted logistic regression, the 'HDL-C/Adiponectin fractions', 'ROM/CRP' and 'leptin/sOB-R' patterns were associated with CRC risk [for the highest quartile vs the lowest, incidence rate ratio (IRR) = 0.69, 95 % CI 0.51-0.93, P-trend = 0.01; IRR = 1.70, 95 % CI 1.30-2.23, P-trend = 0.002; and IRR = 0.79, 95 % CI 0.58-1.07; P-trend = 0.05, respectively]. In contrast, the 'TG/C-peptide' pattern was not associated with CRC risk (IRR = 0.75, 95 % CI 0.56-1.00, P-trend = 0.24). After cases within the first 2 follow-up years were excluded, the 'ROM/CRP' pattern was no longer associated with CRC risk, suggesting potential influence of preclinical disease on these associations. By application of PCA, the study identified 'HDL-C/Adiponectin fractions', 'ROM/CRP' and 'leptin/sOB-R' as biomarker patterns representing potentially important pathways for CRC development.

Biomarker patterns of inflammatory and metabolic pathways are associated with risk of colorectal cancer: Results from the European Prospective Investigation into Cancer and Nutrition (EPIC)

Summary: Azathioprine and cyclophosphamide in doses of 200 mg/day for a proposed duration of one year were compared as adjuncts to prednisolone in the treatment of systemic lupus erythematosus (SLE) in relapse, in a model trial designed to extract maximum information by using multiple indices in few patients. Neither drug proved clearly superior. The possibly greater effectiveness of cyclophosphamide was attributed to a relatively higher effective dosage; however this was offset by more frequent side effects, and in particular the planned 12 months' course of therapy was completed by only one of five cyclophosphamide-treated patients as compared with four of five azathioprine-treated patients. Continuing the trial by including more patients or prolonging treatment was not likely to have yielded additional information. This study on SLE in relapse indicated, at most, that azathioprine and cyclophosphamide may have an accessory role in therapy of this disease, for prednisolone-sparing effects.

A sequential trial comparing cyclophosphamide and azathioprine as adjuncts in the treatment of systemic lupus erythematosus.

A nonparametric feature selection method which can be applicable to pattern recognition problems based on mixed features is presented. In the pattern space, each pattern class is represented by multiple subregions according to local interclass structure. Then in each of the subregions, feature selection is performed in a simple nonparametric way. Our feature selection method can select a feature subset based on higher order discriminating information. Some basic properties of our approach are presented theoretically and experimentally.

Nonparametric Feature Selection Method Based on Local Interclass Structure

The mechanisms underlying caffeine’s acute effects and withdrawal symptoms are not
entirely understood. The purpose was to determine whether these effects or symptoms co-exist in clusters, and whether they are associated with polymorphisms in β1- and β2-adrenoceptors. Subjects (n=1271) were from the Toronto Nutrigenomics and Health Study. The acute effects and withdrawal symptoms clustered into 4 and 6 factors, respectively. Subjects with the ADRβ2 Gly16Arg Gly/Arg genotype were more likely than Gly allele homozygotes to report “fatigue” withdrawal symptoms. Among >200 mg/d caffeine consumers, ADRβ2 Gly allele carriers had a greater risk, compared to Arg allele homozygotes, of reporting ‘flu-like somatic’ withdrawal symptoms. Among subjects with the CYP1A2 -163 A>C A/A genotype and 100-200 mg/d caffeine consumers, ADRβ1 Arg389Gly Gly allele carriers had a greater risk, compared to Arg allele homozygotes, of reporting “dysphoric mood” withdrawal symptoms. The findings suggest that β1- and β2-
adrenoceptors play a role in caffeine withdrawal.%%%%MAST

/pdf/genetic-determinants-of-the-acute-effects-and-withdrawal-58wa3xjo81.pdf

Genetic Determinants of the Acute Effects and Withdrawal Symptoms of Caffeine

Summary: Pattern recognition, which depends upon the perception of inter-relationships between separate observations, has a central role in medical science. Principal component analysis and factor analysis are statistical techniques which can define structural patterns within a set of observations and assign appropriate weights to the importance of each observed variable in contributing to each part of the pattern. The technique of principal component analysis is exemplified by the definition of a “disease activity index” which has been used in the assessment of response of systemic lupus erythematosus to immunosuppressive therapy. Factor analysis of responses to a computer-based questionnaire (SASH) has been used to define patterns of symptoms which correspond to gastrointestinal and cardiac syndromes. Both techniques allow a quantitative approach to the recognition of patterns of disease and should be more widely used in medicine.

V. X. Gledhill

Papers

Pattern Recognition in Medicine — A Place for the Use of Principal Component Analysis and Factor Analysis