Devang Jagdish Doshi MD, Presenter: Nothing to Disclose

Samuel George Armato PhD, Abstract Co-Author: Nothing to Disclose

Roger Engelmann MS, Abstract Co-Author: Nothing to Disclose

Heber MacMahon MD, Abstract Co-Author: Nothing to Disclose

To compare the quality of forward and reverse temporal subtraction (TS) images, in order to increase the yield of clinically acceptable TS images. To quantify the impact of patient positioning differences on TS image quality.

150 pairs of chest radiographs from 72 patients were used. A forward and reverse TS image was created for each pair. Conventionally, the (forward) TS images are created by holding fixed the temporally more recent image and subtracting a warped version of an antecedent image. The reverse TS image is created by holding fixed the antecedent image and subtracting a warped version of the more recent image. Image quality was evaluated subjectively for each lung field of all TS images by an experienced observer on a 1 (very poor) to 5 (excellent) scale. The overall rating for each TS image was the mean of the lung field ratings. Differences in AP inclination and rotation between the temporally sequential images of the same patient were determined through the manual measurement of anatomic landmarks. These differences were correlated with the TS image registration accuracy ratings.

The mean registration accuracy rating assigned to the forward TS images (3.44) was not statistically significantly different than that of the reverse TS images (3.48). However, 12 of 49 (24.5%) of clinically unacceptable forward TS images had a clinically useful reverse TS image. The differences in the mean ratings assigned to the left and right lung fields of the TS images (3.53 and 3.39, respectively) were statistically significant. While large relative out-of-plane positioning differences limit the resulting registration accuracy, small out-of-plane positioning differences do not guarantee good registration.

Production of a reverse TS image can provide an acceptable alternative when the forward TS image is unacceptable. The left lung field tends to have a higher registration rating than the right lung field. The relative AP inclination and rotation between constituent radiographic images limit the quality of registration.

S.G.A.,H.M.: H.M. and S.G.A. have minor equity in R2 Technology, Inc.H.M.: H.M. is also a consultant to Riverain Medical and Median Technologies

To compare the quality of forward and reverse temporal subtraction (TS) images, in order to increase the yield of clinically acceptable TS images. To quantify the impact of patient positioning differences on TS image quality.

150 pairs of chest radiographs from 72 patients were used. A forward and reverse TS image was created for each pair. Conventionally, the (forward) TS images are created by holding fixed the temporally more recent image and subtracting a warped version of an antecedent image. The reverse TS image is created by holding fixed the antecedent image and subtracting a warped version of the more recent image. Image quality was evaluated subjectively for each lung field of all TS images by an experienced observer on a 1 (very poor) to 5 (excellent) scale. The overall rating for each TS image was the mean of the lung field ratings. Differences in AP inclination and rotation between the temporally sequential images of the same patient were determined through the manual measurement of anatomic landmarks. These differences were correlated with the TS image registration accuracy ratings.

The mean registration accuracy rating assigned to the forward TS images (3.44) was not statistically significantly different than that of the reverse TS images (3.48). However, 12 of 49 (24.5%) of clinically unacceptable forward TS images had a clinically useful reverse TS image. The differences in the mean ratings assigned to the left and right lung fields of the TS images (3.53 and 3.39, respectively) were statistically significant. While large relative out-of-plane positioning differences limit the resulting registration accuracy, small out-of-plane positioning differences do not guarantee good registration.

Production of a reverse TS image can provide an acceptable alternative when the forward TS image is unacceptable. The left lung field tends to have a higher registration rating than the right lung field. The relative AP inclination and rotation between constituent radiographic images limit the quality of registration.

S.G.A.,H.M.: H.M. and S.G.A. have minor equity in R2 Technology, Inc.H.M.: H.M. is also a consultant to Riverain Medical and Median Technologies

Doshi, D, Armato, S, Engelmann, R, MacMahon, H, Temporal Subtraction Registration Accuracy in Chest Radiography: Comparison of Forward and Reverse Temporal Subtraction Images and Quantification of the Impact of Patient Positioning Differences on Registration Accuracy.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4408283.html