Purpose: to find a useful decision procedure for the differentiation of obstructive from non-obstructive small bowel (SB) dilatation on the computed tomography (CT).
Materials and methods: four criteria were divided into different degrees and evaluated. These include: (A): continuity (non-continuous, continuous); (B): transition zone (absent, gradual, abrupt); (C): prestenotic SB fluid (minimal, about one-fourth, one-half, and three-fourths, nearly complete); and (D): colonic contents (minimal, moderate, considerable). One hundred fifty-three examinations, 86 obstructive and 67 non-obstructive, were analyzed using chi(2)-square tests to determine the relationship of each criterion to the presence of small bowel obstruction (SBO), whether the proportions of various degrees of these criteria differed significantly among the obstructive and non-obstructive subgroups, and to classify the criteria with a tree-based model (calculated by the computer) for the development of a useful decision procedure.
Results: in each of the four criteria, the trend of probability of obstruction was statistically significant (P=0.0000). The proportions of most, except two, of the various degrees of different criteria in the obstructive and non-obstructive subgroups differed significantly (P<0.01). The obstruction tends to have a continuou dilatation, an abrupt transition, more prestenotic SB fluids, and less colonic contents. The results of classification by a tree-based model were 76 true-positive, ten false-negative, six false-positive, and 61 true-negative. The sensitivity was 88%; specificity was 91%; positive predictive value was 93%; negative predictive value was 86%; and the overall accuracy was 90%.
Conclusion: by analyzing the above four criteria together, a useful tree-based model can be developed and utilized as a supplemental decision procedure for the differentiation of obstructive from non-obstructive SB dilatation. The accuracy can be further promoted if the factor of a recognized pathologic condition is taken into consideration.