"Crack" cocaine-associated stroke.

We present three cases of "crack" cocaine-associated stroke, together with a review of cocaine-associated cerebrovascular complications. Unlike previously reported cases tentatively associating ischemic stroke with cocaine, our patients had no other potential causes for their strokes. Although the exact mechanism of cocaine-related stroke remains uncertain, both disordered neurogenic control of the cerebral circulation as well as systemic factors (ie, acute hypertension) may play a role.

Correlates of tumor size, gender, cell type, and metastasis of resected non-small cell lung cancer and age.

The purpose of this retrospective study was to examine the relationship between tumor volume and age in resected non-small cell lung cancer (NSCLC). Differences exist in the behavior, growth rate, and metastatic potential of solid tumors in both aged humans and experimental animal models. Data from 669 cases of NSCLC resected between 1980 and 1992 were reviewed (445 males; 224 females; median age 67 years, range 16-86). Measurements of the resected tumor in-situ were made in three dimensions, and these were multiplied to give an estimate of the tumor volume. Multiple regression analysis was used to examine the relationship between the tumor volume, age, gender, histological cell type, and TNM nodal score. No direct relationship existed between patient age and tumor volume or nodal score. However, there was a significant relationship between patient gender and tumor volume, i.e., smaller volume tumors in female patients (p = .02). Considering all variables, two relational subgroups were identified: younger male patients with large adenocarcinomas and older female patients with small squamous cell carcinomas (p = .05). We conclude that the relationship between tumor volume and age is complex and dependent on patient gender and tumor cell type.

Prediction of hemorrhagic blood loss with a genetic algorithm neural network.

There is no established method for accurately predicting how much blood loss has occurred during hemorrhage. In the present study, we examine whether a genetic algorithm neural network (GANN) can predict volume of hemorrhage in an experimental model in rats and we compare its accuracy to stepwise linear regression (SLR). Serial measurements of heart period; diastolic, systolic, and mean blood pressures; hemoglobin; pH; arterial PO2; arterial PCO2; bicarbonate; base deficit; and blood loss as percent of total estimated blood volume were made in 33 male Wistar rats during a stepwise hemorrhage. The GANN and SLR used a randomly assigned training set to predict actual volume of hemorrhage in a test set. Diastolic blood pressure, arterial PO2, and base deficit were selected by the GANN as the optimal predictors set. Root mean square error in prediction of estimated blood volume by GANN was significantly lower than by SLR (2.63%, SD 1.44, and 4.22%, SD 3.48, respectively; P < 0.001). A GANN can predict highly accurately and significantly better than SLR volume of hemorrhage without knowledge of prehemorrhage status, rate of blood loss, or trend in physiological variables.

Evolution of artificial neural network architecture: prediction of depression after mania.

Artificial neural networks (ANNs) are compared to standard statistical methods for outcome prediction in biomedical problems. A general method for using genetic algorithms to "evolve" ANN architecture (EANN) is presented. Accuracy of logistic regression, a fully interconnected ANN, and an EANN for predicting depression after mania are examined. All methods showed very good agreement (training set accuracy, chi-square all p < 0.01). However, significant differences were found for stability (test set accuracy); logistic regression being the most unstable and EANN being significantly more stable than a fully interconnected ANN (McNemar p < 0.01). We conclude that the EANN method enhances ANN stability. This approach may have particular relevance for biomedical prediction problems, such as predicting depression after mania.

Comparison of a genetic algorithm neural network with logistic regression for predicting outcome after surgery for patients with nonsmall cell lung carcinoma.

BACKGROUND: Neural networks have been used to predict outcome in cancer patients. Their accuracy compared with standard statistical methods has not been fully assessed. METHODS: In this study, the authors examined the ability to predict the outcome of surgery in 620 patients with nonsmall cell lung carcinoma (NSCLC) by a genetic algorithm neural network (GANN) using Bayes' theorem compared with logistic regression, and the predictive value of tumor volume measures in addition to standard indices such as histologic type and stage. Predictive methods were compared by examining accuracy of classifying target outcome of patients living or dead at 6, 12, 18, and 24 months after surgery. RESULTS: GANN was a significantly better predictor of outcome than logistic regression at all time points (McNemar, P < 0.01). Measures of tumor volume produced significant improvement in the prediction of 12-, 18-, and 24-month time points with GANN, and at 18- and 24-month time points with logistic regression (Wilcoxon matched pairs signed rank test, P < 0.02). CONCLUSIONS: In this study of surgically treated NSCLC patients, outcome predictions were significantly improved by including measures of tumor volume. For predicting individual patient outcome, GANN was found to be highly accurate and significantly better than logistic regression.

'Tumour volume' as a predictor of survival after resection of non-small-cell lung cancer (NSCLC)

Many factors have been individually related to outcome in populations of non-small-cell lung cancer (NSCLC) patients. Factors responsible for the outcome of an individual after surgical resection are poorly understood. We have examined the importance of 'tumour volume' in determining prognosis of patients following resection of NSCLC in a multivariate model. Cox's proportional hazard analysis was used to determine the relative prognostic significance of stage, patient age, gender, tumour cell-type, nodal score and estimated 'tumour volume' in 669 cases with NSCLC treated with surgical resection, of which 280 had died. All factors (except tumour cell-type, P = 0.33) were individually related to survival (P < 0.05). When examined together, survival time was significantly and independently related to 'tumour volume' and stage (P < 0.001), and other factors ceased to be significant. In cases with stage I or II tumours, risk of death was found to increase significantly with increasing estimated 'tumour volume' (23.8% relative increase in hazard to death per doubling of 'tumour volume', 95% confidence interval 13.2-35.2%, P < 0.001 stage I; P < 0.006 stage II). In cases with stage IIIa tumours this factor alone was the significant prognostic variable. In conclusion, an estimate of 'tumour volume' significantly improves prediction of prognosis for individual NSCLC patients with UICC stage I or II tumours.

Clinical features of dementia associated with apolipoprotein epsilon4: discrimination with a neural network genetic algorithm.

BACKGROUND: It is unclear whether the APOE epsilon4 allele is associated with distinct clinical features in dementia. METHOD: 100 cases meeting ICD criteria for dementia were interviewed using standardized instruments and genotyped for APOE. The presence of the epsilon4 allele was used by a genetic algorithm neural network (GANN) to discriminate symptoms and signs. RESULTS: The GANN selected six features: gender, systolic blood pressure, absence of ankle tendon reflexes, history of weight loss, history of falls, and interviewer observed lability of mood. Using these features, a neural network discriminated cases according to epsilon4 highly accurately (area under receiver operating characteristic=0.83, sensitivity=0.78, specificity=0.78). CONCLUSIONS: A GANN is able to discriminate a clinically distinct group of features among dementia patients who express the epsilon4 allele.