Kristin M. Allen

Algorithm for the automated alignment of radiographs for image subtraction

The results of this study indicate that automated alignment of pairs of radiographs produces subtractions that are indistinguishable in quality from subtractions aligned manually by an experienced aligner. In developing the algorithm for automated alignment, care was taken that the criteria used for establishing a window appropriate for testing the quality of alignment were the same for both alignment techniques.

Effects of small angle discrepancies on interpretations of subtraction images

The present study examines interpretations of subtraction images of pairs of radiographs taken at 0 degrees, 1 degree, or 2 degrees of angle discrepancy. The radiographs were taken at each of 48 alveolar crestal sites on 15 dried human skulls. Computer-simulated lesions were induced at the sites on three fourths of the radiographs. Ten instructed dentists were asked to interpret the subtraction images as to the presence or absence of crestal change. A 2-degree angle discrepancy between radiographs resulted in a significant difference in sensitivity from that of radiographic pairs with a 0-degree discrepancy. However, we found no significant difference in sensitivity between 1-degree and 0-degree pairs of radiographs. In conclusion, a 1-degree geometric difference between pairs of radiographs does not significantly contribute to errors in interpretation of subsequent subtraction images.

A model for dentinal caries progression by digital subtraction radiography

Extracted teeth with preexisting carious lesions were incubated in vitro at 37 degrees C in an artificial saliva buffer. Four teeth were exposed to buffer alone, four to buffer containing 5% glucose, and four to buffer containing 5% sucrose. Time-lapse radiographs were made of the carious lesions using an orienting device that fixed the position of the tooth, x-ray film, and x-ray tube collimator. The radiographs were analyzed over a period of 8 weeks using digital subtraction radiography. The subtracted images revealed that three of the four teeth incubated in saliva alone showed an increase in radiodensity (remineralization) at the depths of the lesion; one tooth showed no detectable change. Three of the four teeth incubated in the presence of 5% glucose showed increased radiolucency (demineralization); one showed no detectable change. The four teeth incubated in the presence of 5% sucrose exhibited results similar to those teeth incubated in 5% glucose.

Crestal Alveolar Bone Change in Patients with Periodontitis as Observed by Subtraction Radiography: An Overview

Subtraction radiography is a highly sensitive and useful technique for detecting crestal alveolar bone changes in patients. Utilizing this technique, researchers have demonstrated that 9% of crestal sites lose bone over a six-month period in untreated subjects with periodontitis. On the order of 10-13% of crestal sites were found to lose bone three months post-periodontal therapy which included surgery. Non-surgical therapy resulted in 0.5-2% of crestal sites with bone loss. Subtraction requires radiographs which have closely approximating projection geometry. Presently available technology for taking standardized radiographs based on an occlusal stent system is adequate for obtaining interpretable subtraction images.

Detection of experimentally induced lesions in. subtraction images of cancellous alveolar bone

Cancellous bone lesions of varying sizes were made in specimens of human mandibles and radiographed under geometrically reproducible conditions and observed by viewing of before and after radiographs and subtraction images. Of 178 lesions, traditional comparison of before and after radiographs resulted in 108 correct diagnoses. Analysis of the same lesions by subtraction radiography resulted in 162 (91%) correct diagnoses. This difference in diagnostic accuracy with the use of these two methods of radiographic analyses was significant at the p = 0.0001 level.

Recognition of handwritten foreign mail

Foreign mail recognition (FMR) is part of the more general problem of recognizing destination addresses in a mail stream. It is defined as the problem of finding the country of destination of a mail piece sent to a foreign address. We discuss some of the differences between FMR and domestic mail recognition (DMR) and present its specific challenges. Two complementary baseline algorithms that use heuristics in combining word, character and digit recognizers are presented. Their performance is improved by reducing the search space for the address elements using address configuration distributions. Preliminary results are presented.