Faculty Bibliography

PURPOSE:The aim of this study was to compare qualitative and quantitative image quality and geometric distortion of 4 magnetic resonance diffusion-weighted imaging (DWI) sequences of the prostate using comparable imaging parameters and similar acquisition times. METHODS AND MATERIALS:Axial T2-weighted turbo spin echo images and axial DWI echo-planar imaging (EPI) sequences, including single-shot spin-echo (ss-EPI), readout-segmented multishot (rs-EPI), selective excitation-reduced field of view (sTX-EPI), and prototype single-shot technique applying slice-specific shimming (iShim-EPI) sequences, were acquired at 3 T in 10 healthy volunteers (mean age, 26.1 ± 3.8 years; body mass index, 23.2 ± 3.0 kg/m). Two radiologists, blinded to the type of DWI, independently rated DWIs on a 5-point Likert scale regarding subjective image quality features (resolution, demarcation of prostate capsule, zonal anatomy). Interreader agreement was assessed using the intraclass correlation coefficient. Signal-to-noise ratio (SNR) and apparent diffusion coefficient (ADC) values were assessed separately in the peripheral and transitional zone. For the analysis of geometric distortion, the diameter of the prostate from left to right and from anterior to posterior was measured at the level of the verumontanum on b-1000 DWIs and on T2-weighted turbo spin echo images. Differences were compared using the Wilcoxon rank sum test for qualitative parameters, analysis of variance, and Friedman test for quantitative parameters. A P value of less than 0.05 was considered significant, with correction for multiple comparisons. RESULTS:Interreader agreement was good to excellent (intraclass correlation coefficient, 0.71-0.79) for all qualitative features. Subjective image quality regarding "resolution" was significantly better for ss-EPI than rs-EPI (mean Likert score, 4.25 vs 3.8; P = 0.031) and sTX-EPI (4.25 vs 3.3; P = 0.046) and for iShim-EPI as compared with rs-EPI (4.4 vs 3.8; P = 0.031) and sTX-EPI (4.4 vs 3.3; P = 0.047). There was no significant difference regarding capsule demarcation and zonal anatomy. Signal-to-noise ratio was significantly higher in iShim-EPI than sTX-EPI (SNR ± standard deviation [SD], 28.13 ± 8.21 vs 14.96 ± 2.4; P = 0.015). The ADC values were lower for the peripheral zone in the sTX-EPI than in the ss-EPI (ADC ± SD, 1002.94 ± 83.51 vs 1165.05 ± 115.64; P = 0.013) and the rs-EPI (1002.94 ± 83.51 vs 1244.40 ± 89.95; P = 0.0012) and in the transitional zone in the sTX-EPI compared with the rs-EPI (874.50 ± 200.72 vs 1261.47 ± 179.23; P = 0.0021). There were no statistically significant differences in geometric distortion between all DWI sequences. CONCLUSIONS:Single-shot spin-echo EPI and iShim-EPI showed a tendency toward superior image quality and SNR compared with rs-EPI and sTX-EPI with no significant differences in geometric distortion.

PURPOSE/OBJECTIVE:Intravoxel incoherent motion (IVIM) in diffusion-weighted magnetic resonance imaging (DW-MRI) attributes the signal attenuation to the molecular diffusion and to a faster pseudo-diffusion. Purpose of the study was to demonstrate the feasibility of IVIM for the investigation of intracranial cerebrospinal fluid (CSF) dynamics. METHODS:) along three orthogonal directions (mediolateral (ML), anteroposterior (AP), and craniocaudal (CC)) were acquired during maximum systole and diastole in 10 healthy volunteers (6 males, mean age 36 ± 15 years). A pixel-wise bi-exponential fitting with an iterative nonparametric algorithm was carried out to calculate the following parameters: diffusion coefficient (D), fast diffusion coefficient (D*), and fraction of fast diffusion (f). Region of interest measurements were performed in both lateral ventricles. Comparison of IVIM parameters was performed among two cardiac cycle acquisitions and among the diffusion-encoding directions using a paired Student's t test. RESULTS:f significantly (p < 0.05) depended on the diffusion-encoding direction and on the cardiac cycle (diastole AP 0.30 ± 0.13, ML 0.22 ± 0.12, CC 0.26 ± 0.17; systole AP 0.45 ± 0.17, ML 0.34 ± 0.15, CC 0.40 ± 0.21). Neither a cardiac cycle nor a direction dependency was found among mean D values (which is in line with the expected intraventricular isotropic diffusion) and D* values (p > 0.05 each). CONCLUSION/CONCLUSIONS:The fraction of fast diffusion from IVIM is feasible to detect a direction-dependent and cardiac-dependent pulsatile CSF flow within the lateral ventricles allowing for quantitative monitoring of CSF dynamics. This technique might provide opportunities to further investigate the pathophysiology of various neurological disorders involving altered CSF dynamics.

PURPOSE:The aim of this study was to systematically evaluate the accuracy of quantitative intravoxel incoherent motion (IVIM) analysis of the upper abdomen applying simultaneous multislice (SMS) diffusion-weighted imaging (DWI) to reduce acquisition time. MATERIALS AND METHODS:Diffusion-weighted imaging of parenchymal abdominal organs was performed in 8 healthy volunteers at 3 T using a standard DWI sequence (acceleration factor 1 [AF1]) and an SMS-accelerated echo planar imaging sequence with acceleration factors 2 and 3 (AF2/AF3). Intravoxel incoherent motion analysis was performed with a multistep algorithm for true diffusion coefficient (Dt), pseudodiffusion coefficient (D*), and fraction of perfusion (Fp) measured for the liver, kidney cortex and medulla, pancreas, spleen, and erector spinae muscle. Qualitative and quantitative parameters were compared using a repeated measurement 1-way analysis of variance test and the Bonferroni post hoc method. RESULTS:Simultaneous multislice DWI provided diagnostic image quality in all volunteers with a reduction of scan time of 50% for AF2 (67% for AF3) compared with the standard sequence. Decent IVIM analysis for Dt, D*, and Fp can be calculated on the images of both the SMS sequences AF2 and AF3 with typical organ characteristics of IVIM; however, systematical deviations from AF1 were observed: Dt values increased and Fp decreased significantly with higher acceleration factor for liver, kidney, pancreas, and muscle (P < 0.05). Fitting curves of higher acceleration factors tend to be more monoexponentially shaped. CONCLUSIONS:Simultaneous multislice acceleration provides considerable scan time reduction for upper abdomen DWI with equivalent quality of IVIM analysis compared with the standard nonaccelerated technique. Systematic discrepancies of the true Dt, D*, and Fp for SMS acquisitions need to be considered when comparing to standard DWI sequences.

OBJECTIVE:To train a generic deep learning software (DLS) to classify breast cancer on ultrasound images and to compare its performance to human readers with variable breast imaging experience. METHODS:In this retrospective study, all breast ultrasound examinations from January 1, 2014 to December 31, 2014 at our institution were reviewed. Patients with post-surgical scars, initially indeterminate, or malignant lesions with histological diagnoses or 2-year follow-up were included. The DLS was trained with 70% of the images, and the remaining 30% were used to validate the performance. Three readers with variable expertise also evaluated the validation set (radiologist, resident, medical student). Diagnostic accuracy was assessed with a receiver operating characteristic analysis. RESULTS:82 patients with malignant and 550 with benign lesions were included. Time needed for training was 7 min (DLS). Evaluation time for the test data set were 3.7 s (DLS) and 28, 22 and 25 min for human readers (decreasing experience). Receiver operating characteristic analysis revealed non-significant differences (p-values 0.45-0.47) in the area under the curve of 0.84 (DLS), 0.88 (experienced and intermediate readers) and 0.79 (inexperienced reader). CONCLUSION/CONCLUSIONS:DLS may aid diagnosing cancer on breast ultrasound images with an accuracy comparable to radiologists, and learns better and faster than a human reader with no prior experience. Further clinical trials with dedicated algorithms are warranted. Advances in knowledge: DLS can be trained classify cancer on breast ultrasound images high accuracy even with comparably few training cases. The fast evaluation speed makes real-time image analysis feasible.

PURPOSE/OBJECTIVE:Ga-PSMA-11 PET/MRI might further improve the detection rate in patients with EBR and low PSA values over PET/CT. METHODS:Ga-PSMA-11 PET/MRI for biochemical recurrence in our institution between April and December 2016 with three readers. Median PSA level was 0.99 ng/mL (interquartile range: 3.1 ng/mL). Detection of PSMA-positive lesions within the prostate fossa, local and distant lymph nodes, bones, or visceral organs was recorded. Agreement among observers was evaluated with Fleiss's kappa (k). RESULTS:Overall, in 44 of 56 patients (78.6%) PSMA-positive lesions were detected. In four of nine patients (44.4%) with a PSA < 0.2 ng/mL, suspicious lesions were detected (two pelvic and one paraaortic lymph nodes, and two bone metastases). In eight of 11 patients (72.7%) with a PSA between 0.2 and < 0.5 ng/mL, suspicious lesions were detected (two local recurrences, six lymph nodes, and one bone metastasis). Five out of 20 patients with a PSA < 0.5 ng/mL had extrapelvic disease. In 12 of 15 patients (80.0%) with a PSA between 0.5 and < 2.0 ng/mL, suspicious lesions were detected (four local recurrences, nine lymph nodes, and four bone metastases). In 20 of 21 patients (95.2%) with a PSA >2.0 ng/mL, suspicious lesions were detected. The overall interreader agreement for cancer detection was excellent (κ = 0.796, CI 0.645-0.947). CONCLUSIONS:Ga-PSMA-11 PET/MRI has a high detection rate for recurrent prostate cancer even at very low PSA levels <0.5 ng/mL. Furthermore, even at those low levels extrapelvic disease can be localized in 25% of the cases and local recurrence alone is seen only in 10%.

BACKGROUND:The purpose of this study was to investigate whether any texture features show a correlation with intrahepatic tumor growth before the metastasis is visible to the human eye. METHODS:Eight male C57BL6 mice (age 8-10 weeks) were injected intraportally with syngeneic MC-38 colon cancer cells and two mice were injected with phosphate-buffered saline (sham controls). Small animal magnetic resonance imaging (MRI) at 4.7 T was performed at baseline and days 4, 8, 12, 16, and 20 after injection applying a T2-weighted spin-echo sequence. Texture analysis was performed on the images yielding 32 texture features derived from histogram, gray-level co-occurrence matrix, gray-level run-length matrix, and gray-level size-zone matrix. The features were examined with a linear regression model/Pearson correlation test and hierarchical cluster analysis. From each cluster, the feature with the lowest variance was selected. RESULTS:Tumors were visible on MRI after 20 days. Eighteen features from histogram and the gray-level-matrices exhibited statistically significant correlations before day 20 in the experiment group, but not in the control animals. Cluster analysis revealed three distinct clusters of independent features. The features with the lowest variance were Energy, Short Run Emphasis, and Gray Level Non-Uniformity. CONCLUSIONS:Texture features may quantitatively detect liver metastases before they become visually detectable by the radiologist.

OBJECTIVE:The dual-energy CT (DECT) algorithm for urate detection is feasible only if hyperdense deposits are present. Based on our experience, around half of the performed DECT examinations show no such deposits and thus were useless for this indication. Our diagnostic accuracy study investigates whether conventional radiographs can serve as gatekeeper test prior to DECT for reliable exclusion of such radiopaque deposits. MATERIALS AND METHODS:In this retrospective study, 77 clinically indicated DECT examinations of the hand (n = 29), foot (n = 36) and ankle (n = 12) of 55 patients (13 female, mean age 62±15 years) with suspected gouty arthritis were included. Two blinded readers independently evaluated DECT, gray-scale CT images (reference standard) and corresponding standardized radiographs for the presence/location of dense soft tissue deposits. RESULTS:Interreader agreement for detection of soft tissue deposits with DECT and radiographs was excellent (DECT: both readers, κ = 1; radiographs: both readers, κ = 0.94). DECT showed soft tissue deposits in 54/77 DECT (70%) scans. 30/54 scans (56%) showed deposits on the corresponding radiographs, while in 24 scans (44%) no deposits were seen on radiographs. Test performance of radiographs for soft tissue deposit detection: sensitivity 56%, specificity 100%, PPV 100%, NPV 48.9%, and accuracy 69%. Low density of the deposits was the main reasons for false-negative radiographs (19 cases, 79%), followed by superimposition of deposits by osseous structures (5 cases, 21%). CONCLUSION:Conventional radiographs of the hand, foot and ankle cannot serve as a gatekeeper test for reliable exclusion of radiopaque soft tissue deposits prior to DECT.

PURPOSE/OBJECTIVE:To evaluate the accuracy of a deep learning software (DLS) in the discrimination between phyllodes tumors (PT) and fibroadenomas (FA). METHODS:In this IRB-approved, retrospective, single-center study, we collected all ultrasound images of histologically secured PT (n = 11, 36 images) and a random control group with FA (n = 15, 50 images). The images were analyzed with a DLS designed for industrial grade image analysis, with 33 images withheld from training for validation purposes. The lesions were also interpreted by four radiologists. Diagnostic performance was assessed by the area under the receiver operating characteristic curve (AUC). Sensitivity, specificity, negative and positive predictive values were calculated at the optimal cut-off (Youden Index). RESULTS:The DLS was able to differentiate between PT and FA with good diagnostic accuracy (AUC = 0.73) and high negative predictive value (NPV = 100%). Radiologists showed comparable accuracy (AUC 0.60-0.77) at lower NPV (64-80%). When performing the readout together with the DLS recommendation, the radiologist's accuracy showed a non-significant tendency to improve (AUC 0.75-0.87, p = 0.07). CONCLUSION/CONCLUSIONS:Deep learning based image analysis may be able to exclude PT with a high negative predictive value. Integration into the clinical workflow may enable radiologists to more confidently exclude PT, thereby reducing the number of unnecessary biopsies.