Potential contribution of multiplanar reconstruction (MPR) to computer-aided detection of lung nodules on MDCT

Compressed sensing with deep learning reconstruction: Improving capability of gadolinium-EOB-enhanced 3D T1WI

PURPOSE

The purpose of this study was to determine the utility of compressed sensing (CS) with deep learning reconstruction (DLR) for improving spatial resolution, image quality and focal liver lesion detection on high-resolution contrast-enhanced T1-weighted imaging (HR-CE-T1WI) obtained by CS with DLR as compared with conventional CE-T1WI with parallel imaging (PI).

METHODS

Seventy-seven participants with focal liver lesions underwent conventional CE-T1WI with PI and HR-CE-T1WI, surgical resection, transarterial chemoembolization, and radiofrequency ablation, followed by histopathological or >2-year follow-up examinations in our hospital. Signal-to-noise ratios (SNRs) of liver, spleen and kidney were calculated for each patient, after which each SNR was compared by means of paired t-test. To compare focal lesion detection capabilities of the two methods, a 5-point visual scoring system was adopted for a per lesion basis analysis. Jackknife free-response receiver operating characteristic (JAFROC) analysis was then performed, while sensitivity and false positive rates (/data set) for consensus assessment of the two methods were also compared by using McNemar's test or the signed rank test.

RESULTS

Each SNR of HR-CE-T1WI was significantly higher than that of conventional CE-T1WI with PI (p < 0.05). Sensitivities for consensus assessment showed that HR-CE-MRI had significantly higher sensitivity than conventional CE-T1WI with PI (p = 0.004). Moreover, there were significantly fewer FP/cases for HR-CE-T1WI than for conventional CE-T1WI with PI (p = 0.04).

CONCLUSION

CS with DLR are useful for improving spatial resolution, image quality and focal liver lesion detection capability of Gd-EOB-DTPA enhanced 3D T1WI without any need for longer breath-holding time.

Feasibility of Deep Learning-Based Noise and Artifact Reduction in Coronal Reformation of Contrast-Enhanced Chest Computed Tomography

PURPOSE

This study aimed to evaluate the feasibility of a deep learning method for imaging artifact and noise reduction in coronal reformation of contrast-enhanced chest computed tomography (CT).

METHODS

A total of 19,052 coronal reformatted chest CT images of 110 CT image sets (55 pairs of concordant 16- and 320-row CT image sets) were included and used to train a deep learning algorithm for artifact and noise correction. For internal validation, 4093 coronal reformatted CT images of 25 patients from 16-row CT images underwent correction processing. For external validation, chest CT images of 30 patients (1028 coronal reformatted CT images), acquired in other institutions using different scanners, were subjected to correction processing. For both validations, image quality was compared between original ("CTorigin") and deep learning-based corrected ("CTcorrect") CT images. Quantitative analysis for stair-step artifact (coefficient of variance of CT density on coronal reformation), image noise, signal-to-noise ratio, and contrast-to-noise ratio were evaluated. Subjective image quality scores were assigned for image contrast, artifact, and conspicuity of major structures.

RESULTS

CTcorrect showed significantly reduced stair-step artifact (mean coefficient of variance: CTorigin 7.35 ± 2.0 vs CTcorrect 5.17 ± 2.4, P < 0.001) and image noise and improved signal-to-noise ratio and contrast-to-noise ratio in the aorta, pulmonary artery, and liver, compared with those of CTorigin (P < 0.01). On subjective analysis, CTcorrect had higher image contrast, lower artifact, and better conspicuity than CTorigin. Most results of the external validation were consistent with those obtained from the internal validation, except for those concerning the pulmonary artery.

CONCLUSIONS

Deep learning-based artifact correction significantly improved the image quality of coronal reformation chest CT by reducing image noise and artifacts.

Machine learning for lung CT texture analysis: Improvement of inter-observer agreement for radiological finding classification in patients with pulmonary diseases

PURPOSE

To evaluate the capability ML-based CT texture analysis for improving interobserver agreement and accuracy of radiological finding assessment in patients with COPD, interstitial lung diseases or infectious diseases.

MATERIALS AND METHODS

Training cases (n = 28), validation cases (n = 17) and test cases (n = 89) who underwent thin-section CT at a 320-detector row CT with wide volume scan and two 64-detector row CTs with helical scan were enrolled in this study. From 89 CT data, a total of 350 computationally selected ROI including normal lung, emphysema, nodular lesion, ground-glass opacity, reticulation and honeycomb were evaluated by three radiologists as well as by the software. Inter-observer agreements between consensus reading with and without using the software or software alone and standard references determined by consensus of pulmonologists and chest radiologists were determined using κ statistics. Overall distinguishing accuracies were compared among all methods by McNemar's test.

RESULTS

Agreements for consensus readings obtained with and without the software or the software alone with standard references were determined as significant and substantial or excellent (with the software: κ = 0.91, p < 0.0001; without the software: κ = 0.81, p < 0.0001; the software alone: κ = 0.79, p < 0.0001). Overall differentiation accuracy of consensus reading using the software (94.9 [332/350] %) was significantly higher than that of consensus reading without using the software (84.3 [295/350] %, p < 0.0001) and the software alone (82.3 [288/350] %, p < 0.0001).

CONCLUSION

ML-based CT texture analysis software has potential for improving interobserver agreement and accuracy for radiological finding assessments in patients with COPD, interstitial lung diseases or infectious diseases.

Comparison of computer-aided detection (CADe) capability for pulmonary nodules among standard-, reduced- and ultra-low-dose CTs with and without hybrid type iterative reconstruction technique

PURPOSE

To directly compare the effect of a reconstruction algorithm on nodule detection capability of the computer-aided detection (CADe) system using standard-dose, reduced-dose and ultra-low dose chest CTs with and without adaptive iterative dose reduction 3D (AIDR 3D).

MATERIALS AND METHODS

Our institutional review board approved this study, and written informed consent was obtained from each patient. Standard-, reduced- and ultra-low-dose chest CTs (250 mA, 50 mA and 10 mA) were used to examine 40 patients, 21 males (mean age ± standard deviation: 63.1 ± 11.0 years) and 19 females (mean age, 65.1 ± 12.7 years), and reconstructed as 1 mm-thick sections. Detection of nodule equal to more than 4 mm in dimeter was automatically performed by our proprietary CADe software. The utility of iterative reconstruction method for improving nodule detection capability, sensitivity and false positive rate (/case) of the CADe system using all protocols were compared by means of McNemar's test or signed rank test.

RESULTS

Sensitivity (SE: 0.43) and false-positive rate (FPR: 7.88) of ultra-low-dose CT without AIDR 3D was significantly inferior to those of standard-dose CTs (with AIDR 3D: SE, 0.78, p < .0001, FPR, 3.05, p < .0001; and without AIDR 3D: SE, 0.80, p < .0001, FPR: 2.63, p < .0001), reduced-dose CTs (with AIDR 3D: SE, 0.81, p < .0001, FPR, 3.05, p < .0001; and without AIDR 3D: SE, 0.62, p < .0001, FPR: 2.95, p < .0001) and ultra-low-dose CT with AIDR 3D (SE, 0.79, p < .0001, FPR, 4.88, p = .0001).

CONCLUSION

The AIDR 3D has a significant positive effect on nodule detection capability of the CADe system even when radiation dose is reduced.

A comparison of axial versus coronal image viewing in computer-aided detection of lung nodules on CT

PURPOSE

To compare primarily viewing axial images (Axial mode) versus coronal reconstruction images (Coronal mode) in computer-aided detection (CAD) of lung nodules on multidetector computed tomography (CT) in terms of detection performance and reading time.

MATERIALS AND METHODS

Sixty CT data sets from two institutions were collected prospectively. Ten observers (6 radiologists, 4 pulmonologists) with varying degrees of experience interpreted the data sets using CAD as a second reader (performing nodule detection first without then with aid). The data sets were interpreted twice, once each for Axial and Coronal modes, in two sessions held 4 weeks apart. Jackknife free-response receiver-operating characteristic analysis was used to compare detection performances in the two modes.

RESULTS

Mean figure-of-merit values with and without aid were 0.717 and 0.684 in Axial mode and 0.702 and 0.671 in Coronal mode; use of CAD significantly increased the performance of observers in both modes (P < 0.01). Mean reading times for radiologists did not significantly differ between Axial (156 ± 74 s) and Coronal mode (164 ± 69 s; P = 0.08). Mean reading times for pulmonologists were significantly lower in Coronal (112 ± 53 s) than in Axial mode (130 ± 80 s; P < 0.01).

CONCLUSION

There was no statistically significant difference between Axial and Coronal modes for lung nodule detection with CAD.

Comparison of the effect of iterative reconstruction versus filtered back projection on cardiac CT postprocessing

RATIONALE AND OBJECTIVES

To investigate the impact of iterative reconstruction in image space (IRIS) on image noise, image quality (IQ), and postprocessing at coronary computed tomography angiography (cCTA) compared to traditional filtered back-projection (FBP).

MATERIALS AND METHODS

The cCTA results of 50 patients (26 men; 58 ± 15 years, body mass index 31.5 ± 6.7 kg/m²) were investigated using a second-generation dual-source computed tomography system. Scan data were reconstructed with the use of IRIS and FBP algorithms. Two radiologists independently evaluated the reconstructions using automated coronary tree analysis software. Image noise was measured and IQ was rated on a 5-point Likert scale. The number of manual corrections after automated vessel segmentation, the time required to complete segmentation, and the number of missed segments were assessed in both IRIS and FBP reconstructions. Results were compared using paired t-test.

RESULTS

IRIS significantly reduced image noise compared to FBP (23.3 ± 8.8 vs. 33.5 ± 13.5 Hounsfield units, P < .001). Subjective IQ improved with IRIS (IRIS 3.2 ± 1.0 vs. FBP 3.0 ± 1.0, P < .05). IRIS decreased the time needed for coronary segmentation from 111.9 ± 40.5 seconds to 95.2 ± 38.2 seconds with FBP (P < .01) and required fewer manual corrections (5.7 ± 3.0 vs. 6.8 ± 3.6, P < .01). The number of missed vessel segments was not significantly different (3.6 ± 1.8 vs. 3.8 ± 1.9, P > .05) between IRIS and FBP, respectively.

CONCLUSIONS

During cCTA postprocessing, IRIS significantly decreases the time and the number of manual corrections for a complete coronary segmentation compared to FBP. This effect is likely attributable to suppression of image noise by IRIS, which improves the performance of automated vessel segmentation and positively impacts cCTA analysis.