Statistical analysis of white blood cell data directed to the detection of malignancy in untreated patients

A training set of 90 malignant cases (positives) and 61 non-malignant cases (negatives) was used to determine statistics for five parameters based on white blood cell analysis. The statistics used were thresholds, differences between the positive and negative means, and the weighted correlation matrix. Three decision methods using these statistics were then applied to a test set of 90 positives and 60 negatives. The percentage of correct decisions ranged from 98.67 per cent for parameter 3. The best result using only parameters capable of recognition by an automated apparatus was 94.00 per cent.