Detecting hepatic lesions: the added utility of CT liver window settings

New liver window width in detecting hepatocellular carcinoma on dynamic contrast-enhanced computed tomography with deep learning reconstruction

Changing a window width (WW) alters appearance of noise and contrast of CT images. The aim of this study was to investigate the impact of adjusted WW for deep learning reconstruction (DLR) in detecting hepatocellular carcinomas (HCCs) on CT with DLR. This retrospective study included thirty-five patients who underwent abdominal dynamic contrast-enhanced CT. DLR was used to reconstruct arterial, portal, and delayed phase images. The investigation of the optimal WW involved two blinded readers. Then, five other blinded readers independently read the image sets for detection of HCCs and evaluation of image quality with optimal or conventional liver WW. The optimal WW for detection of HCC was 119 (rounded to 120 in the subsequent analyses) Hounsfield unit (HU), which was the average of adjusted WW in the arterial, portal, and delayed phases. The average figures of merit for the readers for the jackknife alternative free-response receiver operating characteristic analysis to detect HCC were 0.809 (reader 1/2/3/4/5, 0.765/0.798/0.892/0.764/0.827) in the optimal WW (120 HU) and 0.765 (reader 1/2/3/4/5, 0.707/0.769/0.838/0.720/0.791) in the conventional WW (150 HU), and statistically significant difference was observed between them (p < 0.001). Image quality in the optimal WW was superior to those in the conventional WW, and significant difference was seen for some readers (p < 0.041). The optimal WW for detection of HCC was narrower than conventional WW on dynamic contrast-enhanced CT with DLR. Compared with the conventional liver WW, optimal liver WW significantly improved detection performance of HCC.

Interindividual Comparison of Frequency-Selective Nonlinear Blending to Conventional CT for Detection of Focal Liver Lesions Using MRI as the Reference Standard

Background: Diagnosing liver lesions is challenging. CT is used for primary diagnosis; however, its contrast resolution is limited. Investigating methods for better detection of liver lesions is important. Objective: To evaluate the effect of frequency-selective nonlinear blending (NLB) on the detectability of liver lesions on CT. Methods: Retrospective search yielded 109 patients with malignant and benign liver lesions (n = 356; 191 principally diagnosed, 165 incidental findings) underwent contrast-enhanced CT (CECT) in portal venous phase and liver MRI between January 2012 and December 2017. Nonlinear blending (NLB) was applied to CECTs, and three readers rated image quality (5-point Likert scale) in blinded, independent, and randomized fashion. Focal lesions (n = 356) were evaluated for lesion identification and categorization to assess sensitivity. For 191 lesions (primary diagnosis) two readers evaluated CECT and NLB CT to compare size and subjective measurement accuracy. A fourth reader conducted ROI measurements for calculation of contrast-to-noise ratio (CNR), and a fifth reader reviewed MRI as standard of reference. Statistics included interobserver agreement, quantitative comparisons of CNR, lesion size, and subjective image analyses of image quality and sensitivity of detecting liver lesions. Results: Three readers rated image quality of NLB CT higher than for CECT (NLB CT, 4 [10% and 90% percentile: 4, 5]; CECT, 2 [1, 3]; p < .001). CECT showed good interreader agreement (interclass coefficient [ICC], 0.81 [95% CI, 0.76¬-0.85]); NLB CT also (ICC, 0.75 [95% CI, 0.69-0.79]). Mean CNR of liver lesions was increased using NLB (CECT, 4.18 [range, 1.67-9.06]; NLB CT, 12.49 [range, 6.18-23.39]; p < .001). Bland-Altman analysis of lesion size showed reduced underestimation from 2.5 mm (SD, 9.2 [95% CI, 1.2-3.9]) in CECT to 0.1 mm (SD, 3.9 [95% CI, -0.68 to 0.46) for NLB CT (concordance correlation coefficient, 0.99). Sensitivity for detecting liver lesions was increased to 86% for NLB CT (CECT, 76%). Conclusion: Frequency-selective NLB of CECT allows increased image quality and CNR, more precise size measurement, and higher sensitivity for detecting liver lesions. Clinical Impact: NLB CT improves liver lesion detection and increases the accuracy of lesion size measurement, which is important when considering local ablation or liver transplant.

Deep learning-based algorithm to detect primary hepatic malignancy in multiphase CT of patients at high risk for HCC

OBJECTIVES

To develop and evaluate a deep learning-based model capable of detecting primary hepatic malignancies in multiphase CT images of patients at high risk for hepatocellular carcinoma (HCC).

METHODS

A total of 1350 multiphase CT scans of 1280 hepatic malignancies (1202 HCCs and 78 non-HCCs) in 1320 patients at high risk for HCC were retrospectively analyzed. Following the delineation of the focal hepatic lesions according to reference standards, the CT scans were categorized randomly into the training (568 scans), tuning (193 scans), and test (589 scans) sets. Multiphase CT information was subjected to multichannel integration, and livers were automatically segmented before model development. A deep learning-based model capable of detecting malignancies was developed using a mask region-based convolutional neural network. The thresholds of the prediction score and the intersection over union were determined on the tuning set corresponding to the highest sensitivity with < 5 false-positive cases per CT scan. The sensitivity and the number of false-positives of the proposed model on the test set were calculated. Potential causes of false-negatives and false-positives on the test set were analyzed.

RESULTS

This model exhibited a sensitivity of 84.8% with 4.80 false-positives per CT scan on the test set. The most frequent potential causes of false-negatives and false-positives were determined to be atypical enhancement patterns for HCC (71.7%) and registration/segmentation errors (42.7%), respectively.

CONCLUSIONS

The proposed deep learning-based model developed to automatically detect primary hepatic malignancies exhibited an 84.8% of sensitivity with 4.80 false-positives per CT scan in the test set.

KEY POINTS

• Image processing, including multichannel integration of multiphase CT and automatic liver segmentation, enabled the application of a deep learning-based model to detect primary hepatic malignancy. • Our model exhibited a sensitivity of 84.8% with a false-positive rate of 4.80 per CT scan.

Application of Vendor-Neutral Iterative Reconstruction Technique to Pediatric Abdominal Computed Tomography

Objective

To compare image qualities between vendor-neutral and vendor-specific hybrid iterative reconstruction (IR) techniques for abdominopelvic computed tomography (CT) in young patients.

Materials and Methods

In phantom study, we used an anthropomorphic pediatric phantom, age-equivalent to 5-year-old, and reconstructed CT data using traditional filtered back projection (FBP), vendor-specific and vendor-neutral IR techniques (ClariCT; ClariPI) in various radiation doses. Noise, low-contrast detectability and subjective spatial resolution were compared between FBP, vendor-specific (i.e., iDose1 to 5; Philips Healthcare), and vendor-neutral (i.e., ClariCT1 to 5) IR techniques in phantom. In 43 patients (median, 14 years; age range 1–19 years), noise, contrast-to-noise ratio (CNR), and qualitative image quality scores of abdominopelvic CT were compared between FBP, iDose level 4 (iDose4), and ClariCT level 2 (ClariCT2), which showed most similar image quality to clinically used vendor-specific IR images (i.e., iDose4) in phantom study. Noise, CNR, and qualitative imaging scores were compared using one-way repeated measure analysis of variance.

Results

In phantom study, ClariCT2 showed noise level similar to iDose4 (14.68–7.66 Hounsfield unit [HU] vs. 14.78–6.99 HU at CT dose index volume range of 0.8–3.8 mGy). Subjective low-contrast detectability and spatial resolution were similar between ClariCT2 and iDose4. In clinical study, ClariCT2 was equivalent to iDose4 for noise (14.26–17.33 vs. 16.01–18.90) and CNR (3.55–5.24 vs. 3.20–4.60) (p > 0.05). For qualitative imaging scores, the overall image quality ([reader 1, reader 2]; 2.74 vs. 2.07, 3.02 vs. 2.28) and noise (2.88 vs. 2.23, 2.93 vs. 2.33) of ClariCT2 were superior to those of FBP (p < 0.05), and not different from those of iDose4 (2.74 vs. 2.72, 3.02 vs. 2.98; 2.88 vs. 2.77, 2.93 vs. 2.86) (p > 0.05).

Conclusion

Vendor-neutral IR technique shows image quality similar to that of clinically used vendor-specific hybrid IR technique for abdominopelvic CT in young patients.

Imaging Characteristics of Liver Metastases Overlooked at Contrast-Enhanced CT

OBJECTIVE

The purpose of this study was to evaluate the imaging characteristics of liver metastases overlooked at contrast-enhanced CT.

MATERIALS AND METHODS

The records of 746 patients with a diagnosis of liver metastases from colorectal, breast, gastric, or lung cancer between November 2010 and September 2017 were reviewed. Images were reviewed when liver metastases were first diagnosed, and images from prior contrast-enhanced CT examinations were checked if available. These lesions were classified into two groups: missed lesions (those missed on the prior images) and detected lesions (those correctly identified and invisible on the prior images or there were no prior images). Tumor size, contrast-to-noise ratio, location, presence of coexisting liver cysts and hepatic steatosis, and indications for examination were compared between the groups. The t test and Fisher exact test were used to analyze the imaging characteristics of previously overlooked lesions.

RESULTS

The final analysis included 137 lesions, of which 68 were classified as missed. In univariate analysis, contrast-to-noise ratio was significantly lower in missed lesions (95% CI, 2.65 ± 0.24 vs 3.90 ± 0.23; p < 0.001). The proportion of subcapsular lesions (odds ratio, 3.44; p < 0.001), hepatic steatosis (odds ratio, 6.35; p = 0.007), and examination indication other than survey of malignant tumors (odds ratio, 9.07; p = 0.02) were significantly higher for missed lesions.

CONCLUSION

Liver metastases without sufficient contrast enhancement, those in patients with hepatic steatosis, those in subcapsular locations, and those found at examinations for indications other than to assess for tumors were significantly more likely to be overlooked.

Virtual Monoenergetic Images for Diagnostic Assessment of Hypodense Lesions Within the Liver: Semiautomatic Estimation of Window Settings Using Linear Models

OBJECTIVE

The aim of the study was to establish the reference window settings for display of virtual monoenergetic images (VMIs) from spectral detector computed tomography when assessing hypodense liver lesions.

METHODS

In patients with cysts (n = 24) or metastases (n = 26), objective (HU, signal-to-noise ratio [SNR]) and subjective (overall image quality, lesion conspicuity and noise) were assessed. Furthermore, 2 readers determined optimal window center/width (C/W) for conventional images (CIs) and VMIs of 40 to 120 keV. Center/width were modeled against HUliv with and without respect to the keV level (models A and B).

RESULTS

Attenuation and SNR were significantly higher in low-keV VMIs and improved overall image quality and lesion conspicuity (P ≤ 0.05). Model B provided valid estimations of C/W, whereas model A was slightly less accurate.

CONCLUSIONS

The increase in attenuation and SNR on low-keV VMIs requires adjustment of C/W, and they can be estimated in dependency of HUliv using linear models. Reference values for standard display of VMIs of 40 to 120 keV are reported.

Brain computed tomography using iterative reconstruction to diagnose acute middle cerebral artery stroke: usefulness in combination of narrow window setting and thin slice reconstruction

PURPOSE

The purpose of this study is to determine whether iterative model reconstruction (IMR) optimized for brain CT could improve the detection of acute stroke in the setting of thin image slices and narrow window settings.

METHODS

We retrospectively reviewed 27 patients who presented acute middle cerebral artery (MCA) stroke. Images were reconstructed using filtered back projection (FBP; 1- and 5-mm slice thickness) and IMR (1 mm thickness), and contrast-to-noise ratios (CNRs) of infarcted and non-infarcted areas were compared. To analyze the performance of acute MCA stroke detection, we used receiver operating characteristic (ROC) curve techniques and compared 5-mm FBP with standard and narrow window settings, and 1-mm FBP and IMR with narrow window settings.

RESULTS

The CNR in 1-mm IMR (1.1 ± 1.0) was significantly higher than in 5- (0.8 ± 0.7) and 1-mm FBP (0.4 ± 0.4) (p < 0.001). Furthermore, the average area under the ROC curve was significantly higher with 1-mm IMR with narrow window settings (0.90, 95% CI: 0.86, 0.94) than it was with 5-mm FBP (0.78, 95% CI: 0.72, 0.83).

CONCLUSION

The combination of thin image slices and narrow window settings under IMR reconstruction provide better diagnostic performance for acute MCA stroke than conventional reconstruction methods.

Optimizing window settings for improved presentation of virtual monoenergetic images in dual-energy computed tomography

PURPOSE

Dual-energy computed tomography virtual monoenergetic imaging (VMI) at 40 keV exhibits superior contrast-to-noise ratio (CNR), although practicing radiologists do not consistently prefer it over VMI at 70 keV due to high perceivable noise. We hypothesize that the presentation of 40 keV VMI may be compromised using window settings (i.e., window-and-level values [W-L values]) designed for conventional single-energy CT. This study aimed to devise optimum window settings that reduce the apparent noise and utilize the high CNR of 40 keV VMI, in order to improve the conspicuity of hypervascular liver lesions.

MATERIALS AND METHODS

Three W-L value adjustment methods were investigated to alter the presentation of 40 keV VMI. To harness the high CNR of 40 keV VMI, the methods were designed to achieve (a) liver histogram distribution, (b) lesion-to-liver contrast, or (c) liver background noise comparable to those perceived in 70 keV VMI. This IRB-approved study included 18 patient abdominal datasets reconstructed at 40 and 70 keV. For each patient, the W-L values were determined using the three methods. For each of the images with default or adjusted W-L values, the noise, contrast, and CNR were calculated in terms of both display space and native CT number (referred to as HU) space. An observer study was performed to compare the 40 keV images with the three adjusted W-L values, and 40 and 70 keV images with default W-L values in terms of noise, contrast, and diagnostic preference. A comparison was also made in terms of the applicability of using patient-specific or patient-averaged W-L values.

RESULTS

Using the default W-L values, 40 keV VMI exhibited higher HU CNR than 70 keV VMI by 24.6 ± 14.9% (P < 0.001) but lower display CNR by 38.0 ± 16.4% (P < 0.001). Using adjusted W-L values, 40 keV images showed increased display CNR as compared to 70 keV images, by 21.2 ± 13.1%, 17.4 ± 13.6%, and 24.2 ± 15.9% (P < 0.001) for histogram-, noise-, and contrast equalization methods, respectively. The 40 keV images with all three W-L value adjustment methods showed improved perceived conspicuity (CNR) of liver presentation by 103-120% (P < 0.001), as compared to default W-L values. The qualitative observer study revealed that 40 keV images with noise- and histogram-equalized W-L values were the most preferred, followed by 40 keV images with contrast-equalized W-L values and 70 keV images with default W-L values. The 40 keV images with default W-L values were the least preferred. Patient-specific W-L values offered similar results to those of patient-averaged W-L values.

CONCLUSION

The adjusted W-L values can significantly improve the perception of VMI dataset image quality by improving the actual display CNR.

Minimising the impact of errors in the interpretation of CT images for surveillance and evaluation of therapy in cancer

Radiological error is inevitable and usually multifactorial. Error can be secondary to radiologist-specific causes, including cognitive and perceptive errors or ambiguity of report, or system-related causes, including inadequate, misleading, or incorrect clinical information, poor imaging technique, excessive workload, and poor working conditions. In this paper, we discuss a systematic approach to reduce errors in oncological radiology reporting, thus reducing risk to the patient. Rather than attempt to discuss all types of error, we concentrate on the most important and commonly occurring errors that we have encountered over 20 years of practice, based on weekly discrepancy reviews of our practice and independent reviews of clinical and research imaging from other institutions. This review focuses on computed tomography (CT) reporting for staging, surveillance, and response assessment of cancer patients, but the messages apply to all imaging methods.

Knowledge-based iterative model reconstruction: comparative image quality and radiation dose with a pediatric computed tomography phantom

BACKGROUND

CT of pediatric phantoms can provide useful guidance to the optimization of knowledge-based iterative reconstruction CT.

OBJECTIVE

To compare radiation dose and image quality of CT images obtained at different radiation doses reconstructed with knowledge-based iterative reconstruction, hybrid iterative reconstruction and filtered back-projection.

MATERIALS AND METHODS

We scanned a 5-year anthropomorphic phantom at seven levels of radiation. We then reconstructed CT data with knowledge-based iterative reconstruction (iterative model reconstruction [IMR] levels 1, 2 and 3; Philips Healthcare, Andover, MA), hybrid iterative reconstruction (iDose(4), levels 3 and 7; Philips Healthcare, Andover, MA) and filtered back-projection. The noise, signal-to-noise ratio and contrast-to-noise ratio were calculated. We evaluated low-contrast resolutions and detectability by low-contrast targets and subjective and objective spatial resolutions by the line pairs and wire.

RESULTS

With radiation at 100 peak kVp and 100 mAs (3.64 mSv), the relative doses ranged from 5% (0.19 mSv) to 150% (5.46 mSv). Lower noise and higher signal-to-noise, contrast-to-noise and objective spatial resolution were generally achieved in ascending order of filtered back-projection, iDose(4) levels 3 and 7, and IMR levels 1, 2 and 3, at all radiation dose levels. Compared with filtered back-projection at 100% dose, similar noise levels were obtained on IMR level 2 images at 24% dose and iDose(4) level 3 images at 50% dose, respectively. Regarding low-contrast resolution, low-contrast detectability and objective spatial resolution, IMR level 2 images at 24% dose showed comparable image quality with filtered back-projection at 100% dose. Subjective spatial resolution was not greatly affected by reconstruction algorithm.

CONCLUSION

Reduced-dose IMR obtained at 0.92 mSv (24%) showed similar image quality to routine-dose filtered back-projection obtained at 3.64 mSv (100%), and half-dose iDose(4) obtained at 1.81 mSv.

Optimization of window settings for virtual monoenergetic imaging in dual-energy CT of the liver: A multi-reader evaluation of standard monoenergetic and advanced imaged-based monoenergetic datasets

OBJECTIVES

To evaluate optimal window settings for display of virtual monoenergetic reconstructions in third-generation dual-source, dual-energy computed tomography (DECT) of the liver.

METHODS

Twenty-nine subjects were prospectively evaluated with DECT in arterial (AP) and portal venous (PVP) phases. Three reconstructed datasets were calculated: standard linearly-blended (LB120), 70-keV standard virtual monoenergetic (M70), and 50-keV advanced image-based virtual monoenergetic (M50+). Two readers assessed optimal window settings (width and level, W/L), establishing a mean for each reconstruction which was used for a blinded assessment of liver lesions.

RESULTS

The optimal W/L for M50+ were significantly higher for both AP (W=429.3 ± 44.6 HU, L=129.4 ± 9.7 HU) and PVP (W=376.1 ± 14.2HU, L=146.6 ± 7.0 HU) than for LB120 (AP, W=215.9 ± 16.9 HU, L=82.3 ± 9.4 HU) (PVP, W=173.4 ± 8.9 HU, L=69.3 ± 6.0 HU) and M70 (AP, W=247.1 ± 22.2 HU, L=72.9 ± 6.8 HU) (PVP, W=232.0 ± 27.9 HU, L=91.6 ± 14.4 HU). Use of the optimal window setting for M50+ vs. LB120 resulted in higher sensitivity (AP, 100% vs. 86%; PVP, 96% vs. 63%).

CONCLUSIONS

Application of dedicated window settings results in improved liver lesion detection rates in advanced image-based virtual monoenergetic DECT when customized for arterial and portal venous phases.

The importance of surrounding tissues and window settings for contouring of moving targets

AIM

The aim of the study was to assess the importance of surrounding tissues for the delineation of moving targets in tissue-specific phantoms and to find optimal settings for lung, soft tissue, and liver tumors.

MATERIALS AND METHODS

Tumor movement was simulated by a water-filled table tennis ball (target volume, TV). Three phantoms were created: corkboards to simulate lung tissue (lung phantom, LunPh), animal fat as fatty soft tissue (fatty tissue phantom, FatPh), and water enhanced with contrast medium as the liver tissue (liver phantom, LivPh). Slow planning three-dimensional compute tomography images (3D-CTs) were acquired with and without phantom movements. One-dimensional tumor movement (1D), three-dimensional tumor movement (3D), as well as a real patient's tumor trajectories were simulated. The TV was contoured using two lung window settings, two soft-tissue window settings, and one liver window setting. The volumes were compared to mathematical calculated values.

RESULTS

TVs were underestimated in all phantoms due to movement. The use of soft-tissue windows in the LivPh led to a significant underestimation of the TV (70.8% of calculated TV). When common window settings [LunPh + 200 HU/-1,000 HU (upper window/lower window threshold); FatPh: + 240 HU/-120 HU; LivPh: + 175 HU/+ 50 HU] were used, the contoured TVs were: LivPh, 84.0%; LunPh, 93.2%, and FatPh, 92.8%. The lower window threshold had a significant impact on the size of the delineated TV, whereas changes of the upper threshold led only to small differences.

CONCLUSION

The decisive factor for window settings is the lower window threshold (for adequate TV delineation in the lung and fatty-soft tissue it should be lower than density values of surrounding tissue). The use of a liver window should be considered.

Hepatic Parenchymal Heterogeneity on Contrast-enhanced CT Scans Following Oxaliplatin-based Chemotherapy: Natural History and Association with Clinical Evidence of Sinusoidal Obstruction Syndrome

PURPOSE

To evaluate the natural history and contrast material-enhanced computed tomographic (CT) features of postoxaliplatin heterogeneity of liver parenchyma (POHL) and to investigate the association of POHL with clinical factors and biomarkers of sinusoidal obstruction syndrome (SOS).

MATERIALS AND METHODS

The retrospective study was approved by the institutional review board, and informed consent was waived. Two hundred seventy patients (159 men, 111 women; age range, 31-92 years) who underwent oxaliplatin-based chemotherapy (OBC) and serial contrast-enhanced CT were consecutively registered. POHL severity was independently scored by two abdominal imagers, who were blinded to the clinical data, using a six-point scale (POHL presence, ≥4), followed by a consensus review. Complete radiologic remission was determined by consensus on the disappearance of heterogeneity on CT images. The association of POHL severity score with CT-based quantitative (ie, change in spleen size and blood-free hepatic parenchymal attenuation) and laboratory values (ie, aspartate aminotransferase, alanine transaminase, and platelet count), as well as time to complete radiologic remission, were evaluated with the Spearman rank test. Multivariate analysis was performed to determine the association between clinical factors of SOS (ie, age, sex, number of OBC sessions, chemotherapy regimen, bevacizumab use, and presence of concomitant hepatic metastasis) and POHL development. RESULTS Interobserver agreement was excellent (κ = 0.90). POHL was present in 167 (61.9%) of 270 patients, and the number of OBC sessions was significantly associated with POHL development (odds ratio, 1.138; 95% confidence interval: 1.039, 1.245; P = .005). POHL severity score was correlated with quantitative CT and laboratory values (P < .05 for all statistical analysis). Peripheral distribution (103 of 167, 61.7%) and right lobar predominance (103 of 165, 62.4%) were the most common POHL features. The mean time to complete radiologic remission, which was correlated with POHL severity score, was 82.5 days ± 68.8 after OBC discontinuation.

CONCLUSION

POHL development is associated with increased number of OBC sessions, and POHL severity was correlated with various biomarkers of SOS.

Stereotactic body radiotherapy-induced arterial hypervascularity of non-tumorous hepatic parenchyma in patients with hepatocellular carcinoma: potential pitfalls in tumor response evaluation on multiphase computed tomography

Purpose

To evaluate temporal changes in contrast enhancement patterns of non-tumorous hepatic parenchyma with a focus on arterial hypervascularity on multiphase computed tomography (CT) in patients with hepatocellular carcinoma (HCC) treated with stereotactic body radiotherapy (SBRT).

Methods

We retrospectively identified 61 patients who had undergone multiphase contrast-enhanced CT at one, three, and six months after SBRT. Irradiated versus non-irradiated liver parenchyma was delineated by cross-correlation with the dose-volume histogram of SBRT plan. Serial changes in the contrast enhancement patterns of the irradiated versus non-irradiated liver parenchyma were evaluated by two abdominal radiologists in consensus. We compared the frequency of the contrast enhancement patterns according to the follow-up period using the Fisher-Freeman-Halton exact test.

Results

The irradiated non-tumorous hepatic parenchyma showed that the prevalence of arterial hypervascularity increased during the follow-up period (P<.01): 11.5% (7/61) in one, 45.9% (28/61) in three, and 54.1% (33/61) in six months. Contrast wash-out on the delayed phase was uncommon: 1.6% (1/61) in one, 3.3% (2/61) in three, and 0% in six months.

Conclusion

The incidence of arterial hypervascularity of the irradiated hepatic parenchyma gradually increased until six months after SBRT, which could interfere with the accurate evaluation of treatment response. The lack of wash-out on the delayed phase in the hypervascular area would distinguish SBRT-related change from residual/recurred HCC.

Segmentation of hepatic tumor from abdominal CT data using an improved support vector machine framework

An improved support vector machine (SVM) framework has been developed to segment hepatic tumor from CT data. By this method, the one-class SVM (OSVM) and two-class SVM (TSVM) are connected seamlessly by a boosting tool, to tackle the tumor segmentation via both offline and online learning. An initial tumor region was first pre-segmented by an OSVM classifier. Then the boosting tool was employed to automatically generate the negative (non-tumor) samples, according to certain criteria. The pre-segmented initial tumor region and the non-tumor samples generated were used to train a TSVM) classifier. By the trained TSVM classifier, the final tumor lesion was segmented out. Tested on 16 sets of CT abdominal scans, quantitative results suggested that the developed method achieved significantly higher segmentation accuracy than the OSVM and TSVM classifiers.

The value of "liver windows" settings in the detection of small renal cell carcinomas on unenhanced computed tomography

OBJECTIVE

To assess if "liver window" settings improve the conspicuity of small renal cell carcinomas (RCC).

METHODS

Patients were analysed from our institution's pathology-confirmed RCC database that included the following: (1) stage T1a RCCs, (2) an unenhanced computed tomography (CT) abdomen performed ≤ 6 months before histologic diagnosis, and (3) age ≥ 17 years. Patients with multiple tumours, prior nephrectomy, von Hippel-Lindau disease, and polycystic kidney disease were excluded. The unenhanced CT was analysed, and the tumour locations were confirmed by using corresponding contrast-enhanced CT or magnetic resonance imaging studies. Representative single-slice axial, coronal, and sagittal unenhanced CT images were acquired in "soft tissue windows" (width, 400 Hounsfield unit (HU); level, 40 HU) and liver windows (width, 150 HU; level, 88 HU). In addition, single-slice axial, coronal, and sagittal unenhanced CT images of nontumourous renal tissue (obtained from the same cases) were acquired in soft tissue windows and liver windows. These data sets were randomized, unpaired, and were presented independently to 3 blinded radiologists for analysis. The presence or absence of suspicious findings for tumour was scored on a 5-point confidence scale.

RESULTS

Eighty-three of 415 patients met the study criteria. Receiver operating characteristics (ROC) analysis, t test analysis, and kappa analysis were used. ROC analysis showed statistically superior diagnostic performance for liver windows compared with soft tissue windows (area under the curve of 0.923 vs 0.879; P = .0002). Kappa statistics showed "good" vs "moderate" agreement between readers for liver windows compared with soft tissue windows.

CONCLUSION

Use of liver windows settings improves the detection of small RCCs on the unenhanced CT.

Simultaneous presentation of soft tissue and bone tissue in computed tomography with combined window

OBJECTIVES

To examine whether in computed tomography (CT) soft tissue and bone tissue can be simultaneously presented with a combined window without loss of diagnostic information compared to the separate presentations with soft tissue and bone tissue windows.

MATERIALS AND METHODS

Forty-seven CT examinations from different patients with pathological changes at the soft tissue or bone tissue after an accident or due to malignant tumour were evaluated. The CT data were transformed into grey level data with three different windows: (1) soft tissue window; (2) bone tissue window to show spongiosa and bone cortex; and (3) special window to show ethmoidal sinus and mastoidal cells. The images were then weighted with a weighting factor of 2, 3 and 1, respectively, resulting in one image with combined window. This image was compared with the conventional soft tissue and bone tissue images.

RESULTS

All diagnostic information could be obtained and anatomical details be recognized on the image with combined window. In some cases soft tissue structures could be delineated from each other or from adjacent bone better on this image than on the soft tissue image.

CONCLUSIONS

Combined window could enable a reduction of film consumption or digital storage because soft tissue and bone tissue are presented on the same image and not separately. The risk for overlooking important pathological changes might be reduced as both tissues are always presented.

Lesiones focales hepáticas en el mieloma múltiple. Hallazgos en ecografía, tomografía computarizada y resonancia magnética. A propósito de un caso

According to autopsy data, diffuse hepatic infiltration occurs in up to 50% of cases with multiple myeloma. However, focal infiltration (single or multiple), although exceptional, is also possible. Therefore, multiple myeloma should be included in the extensive differential diagnosis of space-occupying liver lesions. We present the case of a 71-year-old man diagnosed with multiple myeloma with multiple focal lesions in the liver and spleen studied by different imaging techniques (ultrasound, multi-phase computed tomography and magnetic resonance). The definitive diagnosis was made by immunohistochemical analysis of material obtained by US-guided fine needle aspiration. We review the imaging findings reported in the literature for focal liver lesions in multiple myeloma and in extramedullary plasmacytoma of the liver.

CT Depiction of Pulmonary Emboli: Display Window Settings

PURPOSE

To compare computed tomographic (CT) window settings selected by radiologists with those determined by using two alternative approaches for depiction of pulmonary emboli (PE).

MATERIALS AND METHODS

Institutional review board approval was obtained; informed consent was not required. This study was compliant with the Health Insurance Portability and Accountability Act. Twenty-five clinical chest CT studies were obtained with a standardized PE protocol and retrospectively evaluated by five chest and two body CT radiologists. Of these studies, 13 were positive for PE, and 12 were negative. At the main pulmonary artery (PA), mean attenuations (MPA) and standard deviations (SDPA) were measured. Initially, images were displayed with a standard mediastinal window setting (window width, W = 400 HU; window center, C = 30 HU), and each observer adjusted the setting to a personally preferred setting (eg, "personal") for PE detection. Images displayed at this setting were compared in a side-by-side fashion with the "modified" (W = MPA + 2 . SDPA, C = W/2) and "double-half" (W = 2 . MPA, C = MPA/2) window setting. Each observer rated images from 1 (ie, most preferred) to 3 (ie, least preferred). For quantitative analysis, window width and center value of each setting were divided by corresponding MPA to compute a width ratio and a center ratio. Window settings and ratings were compared with repeated-measures analysis of variance, paired t tests, and Wilcoxon signed-rank tests.

RESULTS

Ratings for all three types of window settings were significantly different (P < .001). Observers preferred their personal settings the most and the modified settings the least. Mean ratios for the seven observers were 1.68 +/- 0.20 for window width and 0.47 +/- 0.08 for window center. Window width ratios for all settings were significantly different from each other (P < .001). Window center ratios were significantly higher for the modified setting than for the double-half setting (P = .013). Values for mean PA attenuation were correlated with window width ratios for six (86%) observers (mean r2 value = 0.29 +/- 0.19, P < or = .03) and with window center ratios for four (57%) observers (mean r2 value = 0.16 +/- 0.14, P < or = .02), thus indicating a trend of setting window width and window center higher when contrast enhancement is lower and vice versa.

CONCLUSION

On average, observers selected CT window settings for PE detection at a window width of slightly less than twice the mean PA attenuation and at a window center of about half the mean PA attenuation. Observers tended to use larger window widths and centers as the degree of PA enhancement was lower.

A novel display of reconstruction computed tomography for the detection of small hepatocellular carcinoma

PURPOSE

To evaluate the usefulness of the alternate display of arterial and equilibrium phase images (ADAEI) of 2 mm-pitch reconstruction computed tomography (CT) in the detection of hepatocellular carcinoma (HCC).

MATERIALS AND METHODS

One hundred and eleven nodules in 72 patients were confirmed as HCC by radiology, histology, or clinical course. Blinded to the outcome, we retrospectively reviewed the CT images obtained with dual-phase spiral CT (Radix Prima, Hitachi Medical, Tokyo, Japan) by ADAEI and by conventional display on cut films. Scanning for the arterial and equilibrium phases was initiated at 33 and 120 s, respectively, after starting the injection of contrast medium (iopamidol 3 ml/s) with a section thickness of 5 mm and a table feed speed of 5-7 mm/s. In ADAEI, all images were reconstructed with a 2-mm interval, and displayed on the monitor in an alternating fashion so that an image in the arterial phase was followed by the corresponding image in the equilibrium phase, and then by the next pair of images in the craniocaudal direction.

RESULTS

All 20 HCC nodules larger than 20 mm in diameter were detected by both ADAEI and the conventional display (NS). On the other hand, detectability of smaller HCC nodules was 91/91 (100%) and 72/91 (79%), respectively (P<0.0001 by McNemar' test). False-positively identified HCC nodules, including those diagnosed as possible HCC, were 11 by ADAEI and eight by conventional display.

CONCLUSION

The novel, alternate display, ADAEI of 2 mm-pitch reconstruction CT images was useful in detecting small HCC nodules while not requiring additional equipment or expense.

Performance of a single lookup table (LUT) for displaying chest CT images

RATIONALE AND OBJECTIVES

To evaluate the ability to view mediastinal and lung information on one window setting by processing images with a bilinear lookup table (LUT).

MATERIALS AND METHODS

Chest computed tomography (CT) studies were obtained from 32 consecutive adult patient studies, which included 7 iodine contrast studies. From each CT examination, four sections were selected containing the suprasternal notch, carina, right inferior pulmonary vein, and the dome of the lower hemidiaphragm. Each image was processed with a bilinear LUT in addition to the lung setting (window width 1500 and window level -500) and mediastinal setting (window width 450 and window level 50) normally used. Seven radiologists compared the quality of the bilinear LUT with the corresponding lung and mediastinum display settings. A five-point scale was used to assess image quality, with a score of 5 being equivalent to the corresponding lung (or mediastinum) display, 3 being acceptable, and 1 being unacceptable.

RESULTS

The average score of the bilinear LUT for all images and readers was 3.90 +/- 0.93 for lung information and 3.17 +/- 1.00 for mediastinum information. Use of the bilinear LUT resulted in unacceptable images in 0.3% cases for lung information and 5.9% for mediastinum information. Chest CT images that contained iodinated contrast resulted in a higher score than those obtained without contrast, but these differences did not achieve statistical significance.

CONCLUSION

Use of a bilinear LUT has the potential to significantly improve operational efficiency with acceptable image quality for most chest CT images.

Frequency of benign hepatic lesions incidentally detected with contrast-enhanced thin-section portal venous phase spiral CT

PURPOSE

To evaluate the frequency of benign focal hepatic lesions incidentally detected at contrast-enhanced thin-section portal venous phase spiral CT.

MATERIAL AND METHODS

Between January 1998 and February 1999, contrast-enhanced hepatic spiral CT examinations were performed in 1,892 patients. Out of these, only 100 patients fulfilled the following inclusion criteria: No underlying malignant disease, no liver cirrhosis, no suspected or known focal liver lesions. Standardized spiral CT parameters were applied. All CT studies were reviewed retrospectively by one radiologist. Any focal lesion was recorded and classified. Lesion size and number were noted.

RESULTS

A total of 108 hepatic lesions were reported in 33 out of 100 patients (80 cysts; 18 hemangiomas; 3 focal fatty infiltrations; 2 focal non-tumorous perfusion defects; 1 calcification; and 4 non-classified lesions). The average lesion size was 9.4 mm (< or =5 mm: n=40; 6-10 mm: n=30; 11-15 mm: n=28; >15 mm: n=10).

CONCLUSION

Benign liver lesions are probably a frequent incidental finding at abdominal spiral CT.

CT detection of hepatic and splenic injuries: usefulness of liver window settings

OBJECTIVE

This study was designed to assess the usefulness of liver window settings when performing abdominal CT for the detection and characterization of hepatic and splenic injuries.

SUBJECTS AND METHODS

We prospectively evaluated helical abdominal CT scans for hepatic and splenic injuries in 300 consecutive patients with blunt abdominal trauma over a 4-month period. There were 204 males and 96 females with a mean age of 34 years (age range, 1-87 years). For each patient, initial CT diagnosis of hepatic or splenic injury was made from images obtained with standard abdominal window settings. CT scans were then immediately reinterpreted using additional images obtained at narrow window width (liver windows). Changes in conspicuity and characterization of injury were recorded. All CT examinations were performed with helical 7-mm collimation at a pitch of 1.5 after oral ingestion of diluted barium and during bolus IV administration of 125 mL of ioversol at a rate of 2-3 mL/sec.

RESULTS

We detected hepatic or splenic injuries in 34 patients (11.3%). There were 19 hepatic injuries and 18 splenic injuries. Three patients had injuries to both liver and spleen. Conspicuity of hepatic or splenic injuries was mildly increased (+1 H) on liver windows in 16 patients, whereas the injury was equally conspicuous on both liver window and standard window images in 19 cases. In no case did review of the liver windows result in a change in grade of injury or reveal an injury that was not seen on standard abdominal window images. The total increased cost for printing liver windows was $5748.

CONCLUSION

Routine use of liver window settings for abdominal CT in trauma patients has little clinical usefulness and is not cost-effective.

Liver and bone window settings for soft-copy interpretation of chest and abdominal CT

OBJECTIVE

We evaluated whether the use of multiple window and level settings on a soft-copy workstation improves diagnostic accuracy on chest and abdominal CT. We hypothesized that routinely using window and level settings during soft-copy interpretation would beneficially affect the final diagnosis without compromising efficiency.

MATERIALS AND METHODS

Two hundred three randomly selected abdominal and chest CT scans were interpreted by three radiologists using a four-monitor soft-copy workstation (images per screen, nine; resolution, 2K). After the initial interpretations, all scans were reevaluated by the same radiologists using additional liver and bone window and level settings. Differences in conspicuity and characterization of abnormalities were graded on a three-point scale.

RESULTS

Conspicuity and characterization of abnormalities were improved in 67% of abnormal findings (81/121; p = 0.01). Improvement (a finding that substantially affected the final diagnosis) was present in 18% of abnormal findings (22/121; p = 0.04). On average, the evaluation of images at multiple window and level settings required an additional 40 sec per case.

CONCLUSION

The use of multiple window and level settings during soft-copy interpretation resulted in improved lesion detectability and characterization with greater diagnostic efficacy. Using soft-copy workstations, radiologists can evaluate images using multiple settings without compromising efficiency.