Development of an algorithm to automatically compress a CT image to visually lossless threshold

BACKGROUND

To develop an algorithm to predict the visually lossless thresholds (VLTs) of CT images solely using the original images by exploiting the image features and DICOM header information for JPEG2000 compression and to evaluate the algorithm in comparison with pre-existing image fidelity metrics.

METHODS

Five radiologists independently determined the VLT for 206 body CT images for JPEG2000 compression using QUEST procedure. The images were divided into training (n = 103) and testing (n = 103) sets. Using the training set, a multiple linear regression (MLR) model was constructed regarding the image features and DICOM header information as independent variables and regarding the VLTs determined with median value of the radiologists' responses (VLTrad) as dependent variable, after determining an optimal subset of independent variables by backward stepwise selection in a cross-validation scheme. The performance was evaluated on the testing set by measuring absolute differences and intra-class correlation (ICC) coefficient between the VLTrad and the VLTs predicted by the model (VLTmodel). The performance of the model was also compared two metrics, peak signal-to-noise ratio (PSNR) and high-dynamic range visual difference predictor (HDRVDP). The time for computing VLTs between MLR model, PSNR, and HDRVDP were compared using the repeated ANOVA with a post-hoc analysis. P < 0.05 was considered to indicate a statistically significant difference.

RESULTS

The means of absolute differences with the VLTrad were 0.58 (95% CI, 0.48, 0.67), 0.73 (0.61, 0.85), and 0.68 (0.58, 0.79), for the MLR model, PSNR, and HDRVDP, respectively, showing significant difference between them (p < 0.01). The ICC coefficients of MLR model, PSNR, and HDRVDP were 0.88 (95% CI, 0.81, 0.95), 0.85 (0.79, 0.91), and 0.84 (0.77, 0.91). The computing times for calculating VLT per image were 1.5 ± 0.1 s, 3.9 ± 0.3 s, and 68.2 ± 1.4 s, for MLR metric, PSNR, and HDRVDP, respectively.

CONCLUSIONS

The proposed MLR model directly predicting the VLT of a given CT image showed competitive performance to those of image fidelity metrics with less computational expenses. The model would be promising to be used for adaptive compression of CT images.

Computed Tomography Image Compressibility and Limitations of Compression Ratio-Based Guidelines

Finding optimal compression levels for diagnostic imaging is not an easy task. Significant compressibility variations exist between modalities, but little is known about compressibility variations within modalities. Moreover, compressibility is affected by acquisition parameters. In this study, we evaluate the compressibility of thousands of computed tomography (CT) slices acquired with different slice thicknesses, exposures, reconstruction filters, slice collimations, and pitches. We demonstrate that exposure, slice thickness, and reconstruction filters have a significant impact on image compressibility due to an increased high frequency content and a lower acquisition signal-to-noise ratio. We also show that compression ratio is not a good fidelity measure. Therefore, guidelines based on compression ratio should ideally be replaced with other compression measures better correlated with image fidelity. Value-of-interest (VOI) transformations also affect the perception of quality. We have studied the effect of value-of-interest transformation and found significant masking of artifacts when window is widened.

Evaluation of the robustness of the preprocessing technique improving reversible compressibility of CT images: tested on various CT examinations

PURPOSE

To modify the preprocessing technique, which was previously proposed, improving compressibility of computed tomography (CT) images to cover the diversity of three dimensional configurations of different body parts and to evaluate the robustness of the technique in terms of segmentation correctness and increase in reversible compression ratio (CR) for various CT examinations.

METHODS

This study had institutional review board approval with waiver of informed patient consent. A preprocessing technique was previously proposed to improve the compressibility of CT images by replacing pixel values outside the body region with a constant value resulting in maximizing data redundancy. Since the technique was developed aiming at only chest CT images, the authors modified the segmentation method to cover the diversity of three dimensional configurations of different body parts. The modified version was evaluated as follows. In randomly selected 368 CT examinations (352,787 images), each image was preprocessed by using the modified preprocessing technique. Radiologists visually confirmed whether the segmented region covers the body region or not. The images with and without the preprocessing were reversibly compressed using Joint Photographic Experts Group (JPEG), JPEG2000 two-dimensional (2D), and JPEG2000 three-dimensional (3D) compressions. The percentage increase in CR per examination (CRI) was measured.

RESULTS

The rate of correct segmentation was 100.0% (95% CI: 99.9%, 100.0%) for all the examinations. The median of CRI were 26.1% (95% CI: 24.9%, 27.1%), 40.2% (38.5%, 41.1%), and 34.5% (32.7%, 36.2%) in JPEG, JPEG2000 2D, and JPEG2000 3D, respectively.

CONCLUSIONS

In various CT examinations, the modified preprocessing technique can increase in the CR by 25% or more without concerning about degradation of diagnostic information.

The effect of JPEG2000 compression on detection of skull fractures

RATIONAL AND OBJECTIVES

To investigate the effect of the Joint Photographic Experts Group (JPEG2000) 30:1 and 60:1 lossy compression on the detection of cranial vault fractures when compared to JPEG2000 lossless compression.

MATERIALS AND METHODS

Fifty cranial computed tomography (CT) images were processed with three different level of JPEG2000 compression (lossless, 30:1 lossy, and 60:1 lossy) creating three sets of images. These were presented to five musculoskeletal specialists and five neuroradiologists. Each reader read at two of the three compression levels. Twenty-two cases contained a single fracture; the remaining 28 cases contained no fractures. Observers were asked to identify the presence or absence of a fracture, to locate its site, and rate their degree of confidence. Receiver operating characteristic (ROC), jackknife free-response receiver operating characteristic (JAFROC) and the Dorfman-Berbaum-Metz multiple reader multiple case (DBM-MRMC) analyses were used to explore differences between the lossless and lossy compressed images.

RESULTS

JPEG2000 lossless and 30:1 lossy compression demonstrated no significant difference in their performance with JAFROC and DBM-MRMC analysis (P < .416); however, JPEG2000 30:1 lossy compression demonstrated significantly better performance than 60:1 lossy compression (P < .016). A significant increase in misplaced confidence ratings was also seen with 60:1 (P < .037) over 30:1 lossy and lossless compression.

CONCLUSION

JPEG2000 60:1 compression degrades the detection of skull fractures significantly while increasing the confidence with which readers rate fractures compared with 30:1 lossy and lossless compression. JPEG2000 30:1 lossy compression does not significantly change performance when compared to JPEG2000 lossless for the detection of skull fractures on CT.

Use of image features in predicting visually lossless thresholds of JPEG2000 compressed body CT images: initial trial

PURPOSE

To test the image features that may be useful in predicting the visually lossless thresholds (VLTs) of body computed tomographic (CT) images for Joint Photographic Experts Group 2000 (JPEG2000) compression.

MATERIALS AND METHODS

The institutional review board approved this study, with a waiver of informed patient consent. One hundred body CT studies obtained in different patients by using five scanning protocols were obtained, and 100 images, each of which was selected from each of the 100 studies, were collected. Five radiologists independently determined the VLT of each image for JPEG2000 compression by using the QUEST algorithm. The 100 images were randomly divided into two data sets-the training set (50 images) and the testing set (50 images)-and the division was repeated 200 times. For each of the 200 divisions, a multiple linear regression model was constructed on a training set and tested on a testing set regarding each of five image features-standard deviation of image intensity, image entropy, relative percentage of low-frequency (LF) energy, variation in high-frequency (HF) energy, and visual complexity-as independent variables and considering the VLTs determined with the median value of the radiologists' responses as a dependent variable. The root mean square residual and intraclass correlation coefficient (ICC) for the 200 divisions between the VLTs predicted by the models and those determined by radiologists were compared between the models by using repeated-measures analysis of variance with post-hoc comparisons.

RESULTS

Mean root-mean-square residuals for multiple linear regression models constructed with variation in HF energy (1.20 ± 0.10 [standard deviation]) and visual complexity (1.09 ± 0.07) were significantly lower than those for standard deviation of image intensity (1.65 ± 0.13), image entropy (1.63 ± 0.14), and relative percentage of LF energy (1.58 ± 0.12) (P < .01). ICCs for variation in HF energy (0.64 ± 0.05) and visual complexity (0.71 ± 0.04) were significantly higher than those for standard deviation of image intensity (0.04 ± 0.02), image entropy (0.05 ± 0.02), and relative percentage of LF energy (0.20 ± 0.04) (P < .01).

CONCLUSION

Among the five tested image features, variation in HF energy and visual complexity were the most promising in predicting the VLTs of body CT images for JPEG2000 compression.

Predicting the fidelity of JPEG2000 compressed CT images using DICOM header information

PURPOSE

To propose multiple logistic regression (MLR) and artificial neural network (ANN) models constructed using digital imaging and communications in medicine (DICOM) header information in predicting the fidelity of Joint Photographic Experts Group (JPEG) 2000 compressed abdomen computed tomography (CT) images.

METHODS

Our institutional review board approved this study and waived informed patient consent. Using a JPEG2000 algorithm, 360 abdomen CT images were compressed reversibly (n = 48, as negative control) or irreversibly (n = 312) to one of different compression ratios (CRs) ranging from 4:1 to 10:1. Five radiologists independently determined whether the original and compressed images were distinguishable or indistinguishable. The 312 irreversibly compressed images were divided randomly into training (n = 156) and testing (n = 156) sets. The MLR and ANN models were constructed regarding the DICOM header information as independent variables and the pooled radiologists' responses as dependent variable. As independent variables, we selected the CR (DICOM tag number: 0028, 2112), effective tube current-time product (0018, 9332), section thickness (0018, 0050), and field of view (0018, 0090) among the DICOM tags. Using the training set, an optimal subset of independent variables was determined by backward stepwise selection in a four-fold cross-validation scheme. The MLR and ANN models were constructed with the determined independent variables using the training set. The models were then evaluated on the testing set by using receiver-operating-characteristic (ROC) analysis regarding the radiologists' pooled responses as the reference standard and by measuring Spearman rank correlation between the model prediction and the number of radiologists who rated the two images as distinguishable.

RESULTS

The CR and section thickness were determined as the optimal independent variables. The areas under the ROC curve for the MLR and ANN predictions were 0.91 (95% CI; 0.86, 0.95) and 0.92 (0.87, 0.96), respectively. The correlation coefficients of the MLR and ANN predictions with the number of radiologists who responded as distinguishable were 0.76 (0.69, 0.82, p < 0.001) and 0.78 (0.71, 0.83, p < 0.001), respectively.

CONCLUSIONS

The MLR and ANN models constructed using the DICOM header information offer promise in predicting the fidelity of JPEG2000 compressed abdomen CT images.

Advantage in image fidelity and additional computing time of JPEG2000 3D in comparison to JPEG2000 in compressing abdomen CT image datasets of different section thicknesses

PURPOSE

This study aimed to assess the advantage of the Joint Photographic Experts Group 2000 (JPEG2000) 3D (part 2) over JPEG2000 in compressing abdomen computed tomography (CT) image data sets of different section thicknesses (STs).

METHODS

Twenty CT scans were reconstructed with six STs (0.67, 1, 2, 3, 4, and 5 mm) and were then compressed to seven compression ratios (CRs) (reversible, 6:1, 8:1, 10:1, 12:1, 14:1, and 16:1) using JPEG2000 and JPEG2000 3D algorithms. Computing (encoding and decoding) times were measured. The image fidelity of the compressed images was quantitatively measured with two computerized image fidelity metrics, peak signal-to-noise ratio (PSNR) and multiscale structural similarity (MS-SSIM). For 120 selected case-relevant images (20 patients x one image per patient x 6 STs), five radiologists independently compared original and compressed images and assessed the fidelity of the compressed images on a four-grade scale. Wilcoxon signed-rank tests and Friedman tests with post hoc Dunn tests were used for the comparisons between the two compressions and among the six STs, respectively

RESULTS

For each combination of the ST and irreversible CR, JPEG2000 3D showed higher image fidelity than JPEG2000 in terms of PSNR (p < 0.0001), MS-SSIM (p < 0.0001), and five radiologists' grading (p-values ranged from <0.0001 to 0.003). At each CR, the advantage of JPEG2000 3D in image fidelity, measured as the differences in the two computerized image fidelity metrics (PSNR and MS-SSIM), significantly increased as the ST increased from 0.67 to 2 mm, and then slowly decreased as the ST increased from 2 to 5 mm. Similar trends were observed in visual analyses of 120 selected images by five radiologists. At each CR, the 3D-to-2D encoding-time ratio significantly decreased (p < 0.001) as the ST increased from 0.67 to 2 mm, and then slowly increased (p < 0.001) as the ST increased from 2 to 5 mm. The 3D-to-2D decoding-time ratio at each CR did not show a notable biphasic trend across the ST.

CONCLUSIONS

In compressing abdomen CT image data sets of different STs, the advantage of JPEG2000 3D over JPEG2000 increases as the ST increases from 0.67 to 2 mm, and then slowly decreases as the ST increases from 2 to 5 mm. The practical advantage of JPEG2000 3D is limited for a submillimeter ST due to its greater computing time with only a marginal improvement in image fidelity.

A comparison of three image fidelity metrics of different computational principles for JPEG2000 compressed abdomen CT images

This study aimed to evaluate three image fidelity metrics of different computational principles--peak signal-to-noise ratio (PSNR), high-dynamic range visual difference predictor (HDR-VDP), and multiscale structural similarity (MS-SSIM)--in measuring the fidelity of JPEG2000 compressed abdomen computed tomography images from a viewpoint of visually lossless compression. Three hundred images with 0.67- or 5-mm section thickness were compressed to one of five compression ratios ranging from reversible compression to 15:1. The fidelity of each compressed image was measured by five radiologists' visual analyses (distinguishable or indistinguishable from the original) and the three metrics. The Spearman rank correlation coefficients of the PSNR, HDR-VDP, and MS-SSIM values with the number of readers responding as indistinguishable were 0.86, 0.94, and 0.86, respectively. Using the pooled readers' responses as the reference standard, the area under the receiver-operating-characteristic curve for the HDR-VDP (0.99) was significantly greater than that for the PSNR (0.95) (p < 0.001) and for the MS-SSIM (0.96) (p = 0.003), and there was no significant difference between the PSNR and MS-SSIM (p = 0.70). In measuring the image fidelity, the HDR-VDP outperforms the PSNR and MS-SSIM, and the MS-SSIM and PSNR are comparable.