Feature-guided deep learning reduces signal loss and increases lesion CNR in diffusion-weighted imaging of the liver.

PURPOSE: This research aims to develop a feature-guided deep learning approach and compare it with an optimized conventional post-processing algorithm in order to enhance the image quality of diffusion-weighted liver images and, in particular, to reduce the pulsation-induced signal loss occurring predominantly in the left liver lobe. METHODS: Data from 40 patients with liver lesions were used. For the conventional approach, the best-suited out of five examined algorithms was chosen. For the deep learning approach, a U-Net was trained. Instead of learning "gold-standard" target images, the network was trained to optimize four image features (lesion CNR, vessel darkness, data consistency, and pulsation artifact reduction), which could be assessed quantitatively using manually drawn ROIs. A quality score was calculated from these four features. As an additional quality assessment, three radiologists rated different features of the resulting images. RESULTS: The conventional approach could substantially increase the lesion CNR and reduce the pulsation-induced signal loss. However, the vessel darkness was reduced. The deep learning approach increased the lesion CNR and reduced the signal loss to a slightly lower extent, but it could additionally increase the vessel darkness. According to the image quality score, the quality of the deep-learning images was higher than that of the images obtained using the conventional approach. The radiologist ratings were mostly consistent with the quantitative scores, but the overall quality ratings differed among the readers. CONCLUSION: Unlike the conventional algorithm, the deep-learning algorithm increased the vessel darkness. Therefore, it may be a viable alternative to conventional algorithms.

Reduction of the cardiac pulsation artifact and improvement of lesion conspicuity in flow-compensated diffusion images in the liver-A quantitative evaluation of postprocessing algorithms.

PURPOSE: To enhance image quality of flow-compensated diffusion-weighted liver MRI data by increasing the lesion conspicuity and reducing the cardiac pulsation artifact using postprocessing algorithms. METHODS: Diffusion-weighted image data of 40 patients with liver lesions had been acquired at 1.5 T. These data were postprocessed with 5 different algorithms (weighted averaging, p-mean, percentile, outlier exclusion, and exception set). Four image properties of the postprocessed data were evaluated for optimizing the algorithm parameters. These properties were the lesion to tissue contrast-to-noise ratio (CNR), the reduction of the cardiac pulsation artifact, the data consistency, and the vessel darkness. They were combined into a total quality score ( Q total , $$ {Q}_{\mathrm{total}}, $$ set to 1 for the trace-weighted reference image), which was used to rate the image quality objectively. RESULTS: The weighted averaging algorithm performed best according to the total quality score ( Q total = 1.111 ± 0.067 $$ {Q}_{\mathrm{total}}=1.111\pm 0.067 $$ ). The further ranking was outlier exclusion algorithm ( Q total = 1.086 ± 0.061 $$ {Q}_{\mathrm{total}}=1.086\pm 0.061 $$ ), p-mean algorithm ( Q total = 1.045 ± 0.049 $$ {Q}_{\mathrm{total}}=1.045\pm 0.049 $$ ), percentile algorithm ( Q total = 1.012 ± 0.049 $$ {Q}_{\mathrm{total}}=1.012\pm 0.049 $$ ), and exception set algorithm ( Q total = 0.957 ± 0.027 $$ {Q}_{\mathrm{total}}=0.957\pm 0.027 $$ ). All optimized algorithms except for the exception set algorithm corrected the pulsation artifact and increased the lesion CNR. Changes in Q total $$ {Q}_{\mathrm{total}} $$ were significant for all optimized algorithms except for the percentile algorithm. Liver ADC was significantly reduced (except for the exception set algorithm), particularly in the left lobe. CONCLUSION: Postprocessing algorithms should be used for flow-compensated liver DWI. The proposed weighted averaging algorithm seems to be suited best to increase the image quality of artifact-corrupted flow-compensated diffusion-weighted liver data.

Improved Visualization of Prostate Cancer Using Multichannel Computed Diffusion Images: Combining ADC and DWI.

(1) Background: For the peripheral zone of the prostate, diffusion weighted imaging (DWI) is the most important MRI technique; however, a high b-value image (hbDWI) must always be evaluated in conjunction with an apparent diffusion coefficient (ADC) map. We aimed to unify the important contrast features of both a hbDWI and ADC in one single image, termed multichannel computed diffusion images (mcDI), and evaluate the values of these images in a retrospective clinical study; (2) Methods: Based on the 2D histograms of hbDWI and ADC images of 70 patients with histologically proven prostate cancer (PCa) in the peripheral zone, an algorithm was designed to generate the mcDI. Then, three radiologists evaluated the data of 56 other patients twice in three settings (T2w images +): (1) hbDWI and ADC; (2) mcDI; and (3) mcDI, hbDWI, and ADC. The sensitivity, specificity, and inter-reader variability were evaluated; (3) Results: The overall sensitivity/specificity were 0.91/0.78 (hbDWI + ADC), 0.85/0.88 (mcDI), and 0.97/0.88 (mcDI + hbDWI + ADC). The kappa-values for the inter-reader variability were 0.732 (hbDWI + ADC), 0.800 (mcDI), and 0.853 (mcDI + hbDWI + ADC). (4) Conclusions: By using mcDI, the specificity of the MRI detection of PCa was increased at the expense of the sensitivity. By combining the conventional diffusion data with the mcDI data, both the sensitivity and specificity were improved.

Flow-compensated diffusion encoding in MRI for improved liver metastasis detection.

Magnetic resonance (MR) diffusion-weighted imaging (DWI) is often used to detect focal liver lesions (FLLs), though DWI image quality can be limited in the left liver lobe owing to the pulsatile motion of the nearby heart. Flow-compensated (FloCo) diffusion encoding has been shown to reduce this pulsation artifact. The purpose of this prospective study was to intra-individually compare DWI of the liver acquired with conventional monopolar and FloCo diffusion encoding for assessing metastatic FLLs in non-cirrhotic patients. Forty patients with known or suspected multiple metastatic FLLs were included and measured at 1.5 T field strength with a conventional (monopolar) and a FloCo diffusion encoding EPI sequence (single refocused; b-values, 50 and 800 s/mm2). Two board-certified radiologists analyzed the DWI images independently. They issued Likert-scale ratings (1 = worst, 5 = best) for pulsation artifact severity and counted the difference of lesions visible at b = 800 s/mm² separately for small and large FLLs (i.e., < 1 cm or > 1 cm) and separately for left and right liver lobe. Differences between the two diffusion encodings were assessed with the Wilcoxon signed-rank test. Both readers found a reduction in pulsation artifact in the liver with FloCo encoding (p < 0.001 for both liver lobes). More small lesions were detected with FloCo diffusion encoding in both liver lobes (left lobe: six and seven additional lesions by readers 1 and 2, respectively; right lobe: five and seven additional lesions for readers 1 and 2, respectively). Both readers found one additional large lesion in the left liver lobe. Thus, flow-compensated diffusion encoding appears more effective than monopolar diffusion encoding for the detection of liver metastases.

Image Quality and Detection of Small Focal Liver Lesions in Diffusion-Weighted Imaging: Comparison of Navigator Tracking and Free-Breathing Acquisition.

OBJECTIVES: The aim of this study was to compare intraindividual diffusion-weighted imaging (DWI) of the liver acquired with free breathing (FB) versus navigator triggering (NT) for assessing small focal liver lesions (FLLs) in noncirrhotic patients. MATERIALS AND METHODS: Patients with known or suspected multiple FLLs were prospectively included, and spin-echo echo-planar DWI with NT and FB acquisition was performed (b-values, 50 and 800 s/mm2 [b50 and b800]). NT and FB DWI sequences with similar acquisitions times were used. Liver and lesion signal-to-noise ratios were measured at b800. The DWI scans were analyzed independently by 2 readers. Liver edge delineation, presence of stair-step artifacts, vessel sharpness, severity of cardiac motion artifacts, overall image quality, and lesion conspicuity were rated with 5-point Likert scales. Small and large FLLs (ie, <1 cm or ≥1 cm) were rated separately for lesion conspicuity. The FLL detectability was estimated by comparing the number of lesions visible with FB to those visible with NT. RESULTS: Forty-three patients were included in the study. The FB acquisition performed better in terms of severity of cardiac motion artifacts. The NT performed better in terms of liver edge delineation and vessel sharpness. Little difference was found for stair-step artifact, overall image quality, and conspicuity of large FLL, whereas the conspicuity of small FLL was better for NT. For small FLL, both readers found more lesions with NT in 11 cases at b800. For large FLL, this effect was much less pronounced (1 case at b800 reported by 1 of the readers). The mean liver and lesion signal-to-noise ratios were 16.8/41.5 and 19.8/38.4 for NT/FB, respectively. CONCLUSIONS: Small FLL detection is better with NT. Large FLL detection by FB and NT is similarly good. We conclude that NT should be used.

Scanned versus Fused-Reconstructed Oblique MR-Images for Assessment of the Tibiofibular Syndesmosis-Diagnostic PerFormance and Reader Agreement.

To evaluate the diagnostic performance and reader agreement of a novel MRI image fusion method enabling the reconstruction of oblique images for the assessment of the tibiofibular syndesmosis. We evaluated 40 magnetic resonance imaging examinations of patients with ankle sprains (16 with ruptures and 24 without) for the presence of anteroinferior tibiofibular ligament rupture. For all patients, we performed a fusion of standard two-dimensional transversal and coronal 3 mm PDw TSE images into an oblique-fusion reconstruction (OFR) and compared these against conventionally scanned oblique sequence for the evaluation of the tibiofibular syndesmosis. To evaluate diagnostic performance, two expert readers independently read the OFR images twice. We analyzed sensitivity, specificity, negative and positive predictive values, accuracy, and agreement. Reader 1 misinterpreted one OFR as a false negative, demonstrating a sensitivity of 0.94 and specificity of 1.00, reader 2 demonstrated perfect accuracy. Intrareader agreement was almost perfect for reader 1 (α = 0.95) and was perfect for reader 2 (α = 1.00). Additionally, interreader agreement between all fusion sequence reads was almost perfect (α = 0.97). The proposed OFR enables reliable detection of anteroinferior tibiofibular ligament rupture with excellent inter- and intrareader agreement, making conventional scanning of oblique images redundant and supplies a method to retroactively create oblique images, e.g., from external examinations.

On the dependence of the cardiac motion artifact on the breathing cycle in liver diffusion-weighted imaging.

PURPOSE: The purpose of this study was to investigate whether the cardiac motion artifact that regularly appears in diffusion-weighted imaging of the left liver lobe might be reduced by acquiring images in inspiration, when the coupling between heart and liver might be minimal. MATERIALS AND METHODS: 43 patients with known or suspected focal liver lesions were examined at 1.5 T with breath hold acquisition, once in inspiration and once in expiration. Data were acquired with a diffusion-weighted echo planar imaging sequence and two b-values (b50 = 50 s/mm² and b800 = 800 s/mm²). The severity of the cardiac motion artifact in the left liver lobe was rated by two experienced radiologists for both b-values with a 5 point Likert scale. Additionally, the normalized signal S(b800)/S(b50) in the left liver lobe was computed. The Wilcoxon signed-rank test was used comparing the scores of the two readers obtained in inspiration and expiration, and to compare the normalized signal in inspiration and expiration. RESULTS: The normalized signal in inspiration was slightly higher than in expiration (0.349±0.077 vs 0.336±0.058), which would indicate a slight reduction of the cardiac motion artifact, but this difference was not significant (p = 0.24). In the qualitative evaluation, the readers did not observe a significant difference for b50 (reader 1: p = 0.61; reader 2: p = 0.18). For b800, reader 1 observed a significant difference of small effect size favouring expiration (p = 0.03 with a difference of mean Likert scores of 0.27), while reader 2 observed no significant difference (p = 0.62). CONCLUSION: Acquiring the data in inspiration does not lead to a markedly reduced cardiac motion artifact in diffusion-weighted imaging of the left liver lobe and is in this regard not to be preferred over acquiring the data in expiration.

Improved visualization of peripherally inserted central catheters on chest radiographs of neonates using fractional multiscale image processing.

BACKGROUND: Peripherally inserted central catheters (PICCs) provide secure intravenous access for the delivery of life-sustaining medications and nutrition. They are commonly used in pediatrics. Confirmation of correct central catheter tip position is crucial. Verification is usually done by a radiograph. The aim of this study is to evaluate the ability of Fractional Multiscale image Processing (FMP) to detect PICC tips on the digital chest radiographs of neonates. METHODS: A total of 94 radiographs of 47 patients were included in the study. 29 patients were male, 18 were female. The mean age of all examined children was 9.2 days (range 0-99 days). In total, six readers (two radiologists, two residents in radiology, one last year medical student, one neonatologist) evaluated 94 unprocessed and catheter-enhanced radiographs using a 5-point Likert scale (1 = poor catheter tip visualization, 5 = excellent catheter tip visualization). Additionally, the two radiologists evaluated the diagnostic confidence for chest pathologies using a 5-point Likert scale (1 = poor diagnostic confidence, 5 = excellent diagnostic confidence). Radiographs were evaluated on a dedicated workstation. RESULTS: In all cases, the catheter-enhanced radiograph rated higher than (n = 471), or equal (n = 93) to, the unprocessed radiograph when visualizing catheter tips. 87% of the catheter-enhanced radiographs obtained a rating of 4 or higher, while only 42% of unprocessed radiographs received 4 or more points. Regarding diagnostic confidence for chest pathologies one radiologist rated two catheter-enhanced radiographs higher than the unprocessed radiographs, while all other 186 evaluations rated the catheter-enhanced radiographs equal to (n = 78) or lower than (n = 108) the unprocessed radiographs. Only 60% of the catheter-enhanced radiographs yielded a diagnostic confidence of 4 or higher, while 90% of the unprocessed images received 4 or more points. CONCLUSION: Catheter-enhanced digital chest radiographs demonstrate improved visualization of low contrast PICC tips in neonates compared to unprocessed radiographs. Furthermore, they enable detection of accompanying chest pathologies. However, definitive diagnosis of chest pathologies should be made on unprocessed radiographs.

Cardiac T2 star mapping: standardized inline analysis of long and short axis at three identical 1.5 T MRI scanners.

T2 star mapping can be applied for in vivo cardiac iron quantification. Current recommendations of imaging acquisition, post-processing and interpretation of normal values are based on old scanner types and in house software packages. A standardized comparison of short (SAX) and long axis (LAX) segments using commercially available software packages and modern scanners is lacking. To provide a standardized comparison of T2 star time values in SAX and LAX and to investigate intersegmental, interregional and inter-level comparison and the interscanner reproducibility. 84 cardiac MRIs in 28 healthy volunteers were performed with three structurally identical 1.5 T MRI scanners. A commercially available software package for T2 star mapping with automatic in-line motion correction was used for analysis. Regions of interest were manually placed in each of the 16 myocardial segments according to the AHA model in three SAX and three LAX. A total of 2856 ROIs were drawn and 102 segments per volunteer were analysed. Interscanner reproducibility was high (91%) and the mean myocardial T2 star time value for all evaluated segments was 34 ± 5.7 ms. No significant difference was found between all measurements in SAX (35 ± 5.5 ms) and LAX (34 ± 5.8 ms). T2 star time values varied significantly between heart segments in the same axis and in 44% between corresponding SAX and LAX segments. T2 star time values in SAX and LAX have a high interscanner reproducibility but can vary significantly between heart segments in the same axis. Comparability between corresponding SAX and LAX segments is limited. To get representative results T2 star time values should be obtained in more than one heart segment and for follow-up studies identical segments should be used to avoid a systematic bias.

Interobserver and intermodality agreement of standardized algorithms for non-invasive diagnosis of hepatocellular carcinoma in high-risk patients: CEUS-LI-RADS versus MRI-LI-RADS.

OBJECTIVES: We compared the interobserver agreement for the recently introduced contrast-enhanced ultrasound (CEUS)-based algorithm CEUS-LI-RADS (Liver Imaging Reporting and Data System) versus the well-established magnetic resonance imaging (MRI)-LI-RADS for non-invasive diagnosis of hepatocellular carcinoma (HCC) in high-risk patients. METHODS: Focal liver lesions in 50 high-risk patients (mean age 66.2 ± 11.8 years; 39 male) were assessed retrospectively with CEUS and MRI. Two independent observers reviewed CEUS and MRI examinations, separately, classifying observations according to CEUS-LI-RADSv.2016 and MRI-LI-RADSv.2014. Interobserver agreement was assessed with Cohen's kappa. RESULTS: Forty-three lesions were HCCs; two were intrahepatic cholangiocarcinomas; five were benign lesions. Arterial phase hyperenhancement was perceived less frequently with CEUS than with MRI (37/50 / 38/50 lesions = 74%/78% [CEUS; observer 1/observer 2] versus 46/50 / 44/50 lesions = 92%/88% [MRI; observer 1/observer 2]). Washout appearance was observed in 34/50 / 20/50 lesions = 68%/40% with CEUS and 31/50 / 31/50 lesions = 62%/62%) with MRI. Interobserver agreement was moderate for arterial hyperenhancement (ĸ = 0.511/0.565 [CEUS/MRI]) and "washout" (ĸ = 0.490/0.582 [CEUS/MRI]), fair for CEUS-LI-RADS category (ĸ = 0.309) and substantial for MRI-LI-RADS category (ĸ = 0.609). Intermodality agreement was fair for arterial hyperenhancement (ĸ = 0.329), slight to fair for "washout" (ĸ = 0.202) and LI-RADS category (ĸ = 0.218) CONCLUSION: Interobserver agreement is substantial for MRI-LI-RADS and only fair for CEUS-LI-RADS. This is mostly because interobserver agreement in the perception of washout appearance is better in MRI than in CEUS. Further refinement of the LI-RADS algorithms and increasing education and practice may be necessary to improve the concordance between CEUS and MRI for the final LI-RADS categorization. KEY POINTS: • CEUS-LI-RADS and MRI-LIRADS enable standardized non-invasive diagnosis of HCC in high-risk patients. • With CEUS, interobserver agreement is better for arterial hyperenhancement than for "washout". • Interobserver agreement for major features is moderate for both CEUS and MRI. • Interobserver agreement for LI-RADS category is substantial for MRI, and fair for CEUS. • Interobserver-agreement for CEUS-LI-RADS will presumably improve with ongoing use of the algorithm.

Virtual Computed Tomography Colonography: Evaluation of 2D and Virtual 3D Image Quality of Sub-mSv Examinations Enabled by Third-generation Dual Source Scanner Featuring Tin Filtering.

RATIONALE AND OBJECTIVES: To evaluate two- and three-dimensional (2D and 3D) image quality of sub-milliSievert (mSv) computed tomography (CT) colonography utilizing a third-generation dual source CT scanner featuring a tin filter. METHODS: We retrospectively evaluated 26 consecutive patients who underwent third-generation dual source CT colonography, nine with the standard-dose clinical-scan protocol (SDP) and 17 with a low-dose protocol (LDP) featuring a tin filter. Radiation dose was evaluated by volume computed tomography dose index (CTDIvol), dose length product (DLP), effective dose (E), and size-specific dose estimate. Objective image quality was evaluated utilizing signal-to-noise ratio (SNR) derived from standardized placed regions of interest on the transverse 2D images and the ratio of SNR/CTDIvol (normalized SNR). Two radiologists in consensus assessed subjective image quality of the virtual 3D images. RESULTS: There were no significant differences in subjective image quality (P = .661). All examinations were rated "excellent" or "good" for diagnostic confidence. The mean total for DLP/E was 143.4 ± 29.8 mGy/3.00 ± 0.40 mSv in the SDP and therefore significantly higher than in the LDP with 36.9 ± 8.7 mGy/0.75 ± 0.16 mSv (P < .001). The SNR was 8.9 ± 2.1 in the SDP and 4.9 ± 0.8 in the LDP. CONCLUSIONS: Third-generation dual source CT featuring a tin filter enables consistent sub-mSv colonography without substantially impairing image quality.

Vascularization of the Arteriovenous Loop in a Rat Isolation Chamber Model-Quantification of Hypoxia and Evaluation of Its Effects.

INTRODUCTION: The aim of this study was to analyze the three-dimensional distribution of hypoxia in the arteriovenous (AV) loop model in rats, by examining the distribution of hypoxia-inducible factor-1 alpha (HIF-1α). MATERIALS AND METHODS: AV loops were created from the femoral artery and vein of male Lewis rats and an interpositional graft from the contralateral femoral vein. This AV fistula was embedded in a fibrin-filled isolation chamber and subcutaneously implanted into the thigh. The specimens were harvested after 7 days (n = 4), 10 days (n = 5), and 14 days (n = 4). The fibrin clots were stained for lectin, HIF-1α, and ectodysplasin 1 (ED1). The distribution of positive and negative cells was analyzed in three dimensions and at different points in time. RESULTS: The HIF-1α-positive rate increased from the proximity of the central vessel to the distant regions. From day 7 to 10, we noted a decrease in the HIF-1α-positive rate in the proximity of the vessels and an increase in the periphery. A global decrease in positive cells was seen at day 14. HIF-1α and macrophage (ED1) double staining indicated that macrophages accounted for a significant fraction of the cells. Double staining for endothelium (with lectin) demonstrated that no HIF-1α was detectable in well-vascularized areas. CONCLUSION: In the AV loop model, the HIF-1α-positive cell distribution is highly related to the vascularization process. The onset of rapid vessel outgrowth follows the increase of the HIF-1α rate closely, indicating that HIF-1α may be a driving force for vascularization.

Accuracy and Specific Value of Cardiovascular 3D-Models in Pediatric CT-Angiography.

Computed tomography (CT)-angiography is routinely performed prior to catheter-based and surgical treatment in congenital heart disease. To date, little is known about the accuracy and advantage of different 3D-reconstructions in CT-data. Exact anatomical information is crucial. We analyzed 35 consecutive CT-angiographies of infants with congenital heart disease. All datasets are reconstructed three-dimensionally using volume rendering technique (VRT) and threshold-based segmentation (stereolithographic model, STL). Additionally, the two-dimensional maximum intensity projection (MIP) reconstructs two-dimensional data. In each dataset and resulting image, measurements of vascular diameters for four different vessels were estimated and compared to the reference standard, measured via multiplanar reformation (MPR). The resulting measurements obtained via the STL-images, MIP-images, and the VRT-images were compared with the reference standard. There was a significant difference (p < 0.05) between measurements. The mean difference was 0.0 for STL-images, -0.1 for MIP-images, and -0.3 for VRT-images. The range of the differences was -0.7 to 1.0 mm for STL-images, -0.6 to 0.5 mm for MIP-images and -1.1 to 0.7 mm for VRT-images. There was an excellent correlation between the STL-, MIP-, VRT-measurements, and the reference standard. Inter-reader reliability was excellent (p < 0.01). STL-models of cardiovascular structures are more accurate than the traditional VRT-models. Additionally, they can be standardized and are reproducible.

3 Tesla 23Na Magnetic Resonance Imaging During Acute Kidney Injury.

RATIONALE AND OBJECTIVES: Sodium and proton magnetic resonance imaging (23Na/1H-MRI) have shown that muscle and skin can store Na+ without water. In chronic renal failure and in heart failure, Na+ mobilization occurs, but is variable depending on age, dialysis vintage, and other features. Na+ storage depots have not been studied in patients with acute kidney injury (AKI). MATERIALS AND METHODS: We studied 7 patients with AKI (mean age: 51.7 years; range: 25-84) and 14 age-matched and gender-matched healthy controls. All underwent 23Na/1H-MRI at the calf. Patients were studied before and after acute hemodialysis therapy within 5-6 days. The 23Na-MRI produced grayscale images containing Na+ phantoms, which served to quantify Na+ contents. A fat-suppressed inversion recovery sequence was used to quantify H2O content. RESULTS: Plasma Na+ levels did not change. Mean Na+ contents in muscle and skin did not significantly change following four to five cycles of hemodialysis treatment (before therapy: 32.7 ± 6.9 and 44.2 ± 13.5 mmol/L, respectively; after dialysis: 31.7 ± 10.2 and 42.8 ± 11.8 mmol/L, respectively; P > .05). Water content measurements did not differ significantly before and after hemodialysis in muscle and skin (P > .05). Na+ contents in calf muscle and skin of patients before hemodialysis were significantly higher than in healthy subjects (16.6 ± 2.1 and 17.9 ± 3.2) and remained significantly elevated after hemodialysis. CONCLUSIONS: Na+ in muscle and skin accumulates in patients with AKI and, in contrast to patients receiving chronic hemodialysis and those with acute heart failure, is not mobilized with hemodialysis within 5-6 days.

Semi-automated De-identification of German Content Sensitive Reports for Big Data Analytics.

Purpose Projects involving collaborations between different institutions require data security via selective de-identification of words or phrases. A semi-automated de-identification tool was developed and evaluated on different types of medical reports natively and after adapting the algorithm to the text structure. Materials and Methods A semi-automated de-identification tool was developed and evaluated for its sensitivity and specificity in detecting sensitive content in written reports. Data from 4671 pathology reports (4105 + 566 in two different formats), 2804 medical reports, 1008 operation reports, and 6223 radiology reports of 1167 patients suffering from breast cancer were de-identified. The content was itemized into four categories: direct identifiers (name, address), indirect identifiers (date of birth/operation, medical ID, etc.), medical terms, and filler words. The software was tested natively (without training) in order to establish a baseline. The reports were manually edited and the model re-trained for the next test set. After manually editing 25, 50, 100, 250, 500 and if applicable 1000 reports of each type re-training was applied. Results In the native test, 61.3 % of direct and 80.8 % of the indirect identifiers were detected. The performance (P) increased to 91.4 % (P25), 96.7 % (P50), 99.5 % (P100), 99.6 % (P250), 99.7 % (P500) and 100 % (P1000) for direct identifiers and to 93.2 % (P25), 97.9 % (P50), 97.2 % (P100), 98.9 % (P250), 99.0 % (P500) and 99.3 % (P1000) for indirect identifiers. Without training, 5.3 % of medical terms were falsely flagged as critical data. The performance increased, after training, to 4.0 % (P25), 3.6 % (P50), 4.0 % (P100), 3.7 % (P250), 4.3 % (P500), and 3.1 % (P1000). Roughly 0.1 % of filler words were falsely flagged. Conclusion Training of the developed de-identification tool continuously improved its performance. Training with roughly 100 edited reports enables reliable detection and labeling of sensitive data in different types of medical reports. Key Points: · Collaborations between different institutions require de-identification of patients' data. · Software-based de-identification of content-sensitive reports grows in importance as a result of 'Big data'. · A de-identification software was developed and tested natively and after training. · The proposed de-identification software worked quite reliably, following training with roughly 100 edited reports. · A final check of the texts by an authorized person remains necessary. Citation Format · Seuss H, Dankerl P, Ihle M et al. Semi-automated De-identification of German Content Sensitive Reports for Big Data Analytics. Fortschr Röntgenstr 2017; 189: 661 - 671.

Evaluation of Rib Fractures on a Single-in-plane Image Reformation of the Rib Cage in CT Examinations.

RATIONALE AND OBJECTIVES: This study aimed to evaluate the diagnostic performance of using a reformatted single-in-plane image reformation of the rib cage for the detection of rib fractures in computed tomography (CT) examinations, employing different levels of radiological experience. MATERIALS AND METHODS: We retrospectively evaluated 10 consecutive patients with and 10 patients without rib fractures, whose CT scans were reformatted to a single-in-plane image reformation of the rib cage. Eight readers (two radiologists, two residents in radiology, and four interns) independently evaluated the images for the presence of rib fractures using a reformatted single-in-plane image and a multi-planar image reformation. The time limit was 30 seconds for each read. A consensus of two radiologist readings was considered as the reference standard. Diagnostic performance (sensitivity, specificity, positive predictive value [PPV], and negative predictive value [NPV]) was assessed and evaluated per rib and per location (anterior, lateral, posterior). To determine the time limit, we prospectively analyzed the average time it took radiologists to assess the rib cage, in a bone window setting, in 50 routine CT examinations. McNemar test was used to compare the diagnostic performances. RESULTS: Single image reformation was successful in all 20 patients. The sensitivity, specificity, PPV, and NPV for the detection of rib fractures using the conventional multi-planar read were 77.5%, 99.2%, 89.9%, and 98.0% for radiologists; 46.3%, 99.7%, 92.5%, and 95.3% for residents; and 29.4%, 99.4%, 82.5%, and 93.9% for interns, respectively. Sensitivity, PPV, and NPV increased across all three groups of experience, using the reformatted single-in-plane image of the rib cage (radiologists: 85.0%, 98.6%, and 98.7%; residents: 80.0%, 92.8%, and 98.2%; interns: 66.9%, 89.9%, and 97.1%), whereas specificity did not change significantly (99.9%, 99.4%, and 99.3%). The diagnostic performance of the interns and residents was significantly better when evaluating the single-in-plane image reformations (P < .01). The diagnostic performance of the radiologists was better when evaluating the single-in-plane image reformations; however, there was no significant difference (statistical power: 0.32). CONCLUSIONS: The diagnostic performance for the detection of rib fractures, using CT images that have been reformatted to a single-in-plane image, improves for readers from different educational levels when the evaluation time is restricted to 30 seconds or less.

Computer-aided evaluation of low-dose and low-contrast agent third-generation dual-source CT angiography prior to transcatheter aortic valve implantation (TAVI).

PURPOSE: To evaluate the performance of computer-aided evaluation software for a comprehensive workup of patients prior to transcatheter aortic valve implantation (TAVI) using low-contrast agent and low radiation dose third-generation dual-source CT angiography. METHODS: We evaluated 30 consecutive patients scheduled for TAVI. All patients underwent ECG-triggered high-pitch dual-source CT angiography of the aortic root and aorta with a standardized contrast agent volume (30 ml Imeron350, flow rate 4 ml/s) and low-dose (100 kv/350 mAs) protocol. An expert (10 years of experience) manually evaluated aortic root and iliac access dimensions (distance between coronary ostia and aortic annulus, minimal/maximal diameters and area-derived diameter of the aortic annulus) and best CT-predicted fluoroscopic projection angle as the reference standard. Utilizing computer-aided software (syngo.via), the same pre-TAVI workup was performed and compared to the reference standard. RESULTS: Mean CTDI[Formula: see text] was 3.46 mGy and mean DLP 217.6 ± 12.1 mGy cm, corresponding to a mean effective dose of 3.7 ± 0.2 mSv. Computer-aided evaluation was successful in all but one patient. Compared to the reference standard, Bland-Altman analysis indicated very good agreement for the distances between aortic annulus and coronary ostia (RCA: mean difference 0.8 mm; 95 % CI 0.4-1.2 mm; LM: mean difference 0.9 mm; 95 % CI 0.5-1.3 mm); however, we demonstrated a systematic overestimation of annulus- derived diameter using the software (mean difference 44.4 mm[Formula: see text]; 95 % CI 30.4-58.3 mm[Formula: see text]). Based on respective annulus dimensions, the recommended prosthesis size (Edwards SAPIEN 3) matched in 26 out of the 29 patients (90 %). CT-derived fluoroscopic projection angles showed an excellent agreement for both methods. Out of 58 iliac arteries, 15 (25 %) arteries could not be segmented by the software. Preprocessing time of the software was 71 ± 11 s (range 51-96 s), and reading time with the software was 118 ± 31 s (range 68-201 s). CONCLUSION: In the workup of pre-TAVI CT angiography, computer-aided evaluation of low-contrast, low-dose examinations is feasible with good agreement and quick reading time. However, a systematic overestimation of the aortic annulus area is observed.

Development and Evaluation of a Semi-automated Segmentation Tool and a Modified Ellipsoid Formula for Volumetric Analysis of the Kidney in Non-contrast T2-Weighted MR Images.

Volumetric analysis of the kidney parenchyma provides additional information for the detection and monitoring of various renal diseases. Therefore the purposes of the study were to develop and evaluate a semi-automated segmentation tool and a modified ellipsoid formula for volumetric analysis of the kidney in non-contrast T2-weighted magnetic resonance (MR)-images. Three readers performed semi-automated segmentation of the total kidney volume (TKV) in axial, non-contrast-enhanced T2-weighted MR-images of 24 healthy volunteers (48 kidneys) twice. A semi-automated threshold-based segmentation tool was developed to segment the kidney parenchyma. Furthermore, the three readers measured renal dimensions (length, width, depth) and applied different formulas to calculate the TKV. Manual segmentation served as a reference volume. Volumes of the different methods were compared and time required was recorded. There was no significant difference between the semi-automatically and manually segmented TKV (p = 0.31). The difference in mean volumes was 0.3 ml (95% confidence interval (CI), -10.1 to 10.7 ml). Semi-automated segmentation was significantly faster than manual segmentation, with a mean difference = 188 s (220 vs. 408 s); p < 0.05. Volumes did not differ significantly comparing the results of different readers. Calculation of TKV with a modified ellipsoid formula (ellipsoid volume × 0.85) did not differ significantly from the reference volume; however, the mean error was three times higher (difference of mean volumes -0.1 ml; CI -31.1 to 30.9 ml; p = 0.95). Applying the modified ellipsoid formula was the fastest way to get an estimation of the renal volume (41 s). Semi-automated segmentation and volumetric analysis of the kidney in native T2-weighted MR data delivers accurate and reproducible results and was significantly faster than manual segmentation. Applying a modified ellipsoid formula quickly provides an accurate kidney volume.