[Application of fuzzy inference to personal identification--sex determination from teeth]

Fuzzy inference has received much attention recently as a new type of computer control technology for application in various fields. In this study I applied fuzzy inference to personal identification in medical jurisprudence. Sex determination program was constructed from teeth using the fuzzy inference technique and was compared with known discriminant analysis. The materials examined were mandibular and maxillary dental plaster models of 100 adult males and females, and 80 infant males and females. For the permanent teeth, measurements were made on crown width and thickness of the mandibular and maxillary central incisors, canines, first premolars, and first molars, and of the width and length of the maxillary dental arch. For the deciduous teeth, measurements were made on crown width and thickness of all teeth. Stepwise discriminant analysis was conducted using these values. The values from 5 selected upper-ranked items were set as input objects of the fuzzy inference program and the probability of maleness obtained from the discriminant analysis was set as the output object. Finally, sex determination program for both permanent and deciduous teeth was constructed using a fuzzy inference software development tool. Each measured value was input into this program and the output results were compared with those of discriminant analysis. The percentages of correct determinations for permanent teeth were 83.0% for males and 86.0% for females using stepwise discriminant analysis, however it increased to 93.0% for males and 89.0% for females using the fuzzy inference program. The percentages of correct determinations for deciduous teeth models were 67.5% for males and 75.0% for females using stepwise discriminant analysis, and increased to 86.3% for males and 81.3% for females using the fuzzy inference program. Among samples with probabilities between 40% and 60%, 3 out of 14 males and 4 out of 13 females were misjudged using stepwise discriminant analysis in cases of permanent teeth. However using fuzzy inference program, it reduced to 0 for males and 3 for females. In case of deciduous teeth, 9 out of 20 males and 9 out of 23 females were misjudged using stepwise discriminant analysis. But it reduced to 5 for males and 4 for females using the fuzzy inference program.