Diagnosis of Retroperitoneal Masses by Gastrointestinal Roentgenographs Measurements: a Computer Study

Abstract

Many authors (1–8) have tried to establish normal ranges for evaluating the distance from stomach and duo-denum to spinal and abdominal landmarks. The span between these organs and the spine has been of special interest because enlargement of the prevertebral space by retroperitoneal masses can be observed on gastrointestinal roentgenograms. No reliable limits of normal measurements have been found which would help to identify a retrogastric or retroduodenal mass in an individual case. A diagnosis of retroperitoneal tumor suspected only because of a large distance between gut and spine, even without tumor “impression,” is frequently proved correct. Unrecognized combinations of observations and subjective factors may add to the suspicion of the observer. A comprehensive analysis of nine measurements and of six possible subjective factors was undertaken in 200 normal subjects and 72 tumor patients. The objective was to increase the possibility of discovering worthwhile factors or combinations of factors now ignored, and to better evaluate the worth of gastrointestinal roentgenographic measurements as an aid to diagnosis of retroperitoneal tumor. Method Table I and Figures 1 and 2 illustrate the measurements and observations made. Extensive tests with roentgenograms obtained in various positions before the study was instituted proved the erect position to be the most conducive to reliable measurements. Short, 120-kv exposures permitted moving and overlapping viscera to be seen. Measurements and observations were coded on punch cards. A 7090 IBM computer at the University of California at Berkeley and a 1620 IBM computer at San Francisco were used. The statistician employed the FORTRAN method of programming to obtain measurements and ratio ranges, standard deviations, and averages. These computations were also subgrouped and reanalyzed for correlation with age, height, weight, body surface, filling of the stomach, tumor “impression,” and pathologic changes in proved tumors (Table II). Sex differences were not analyzed, as spot-checking suggested little correlation aside from the factors of body build under investigation. The power of each factor and combination of factors to separate normal from abnormal was computed by “discriminate function.”2 Combinations of variables were studied in “distribution analysis” graphs made by the IBM plotter. Material Calculations were made from 272 gastrointestinal studies, 200 normal and 72 with retrogastric or retroduodenal masses. To obtain the normal measurements, all gastrointestinal series included erect, anteroposterior, and lateral views until 200 normal studies were collected. The retroperitoneal tumor cases were gathered from previous roentgenograms obtained without uniformity in technic except that all consisted of anteroposterior and lateral views. In this retrospective series erect roentgenograms were not available in all the patients with tumors.