Accuracy For Detection Of Lesions Research Articles

Computer Aided Detection of Breast Cancer on Ultrasound Imaging Using Deep Learning

Although breast ultrasound imaging is powerful and effective tool to detect breast lesions and have been widely performed worldwide, it is an operator-dependent test, hence the accuracy for detection and diagnosis of breast lesions depend on the operator. We develop a computer-aided detection system for masses in ultrasound image using deep convolutional neural network. A data set containing 818 ultrasound images were collected from our institute. Breast lesions on these images were cropped as 256x256 pixel images manually. All images were flipped horizontally and were collected to make new images. After data augmentation, a total of 2604 images were obtained. These images were labeled and classified by histological tissue type: cyst, concentrated cyst, ductal carcinoma in situ, fibroadenoma, intraductal papilloma, lymph node, invasive lobular carcinoma, mastitis, mastopathic change, mucinous carcinoma, invasive papillotubular carcinoma, phyllodes, tumor, invasive scirrhous carcinoma or invasive solidtubular carcinoma. For training and testing, the data set was randomly divided into a training set and an independent test set with a ratio of 80:20. Classification of data set images were trained using convolutional neural network(CNN) with ten hidden layers including convolutional layers and pooling layers. Its accuracy of classification was evaluated. Data set was also classified as either benign or malignant, and trained using CNN. Its accuracy, sensitivity and specificity were also evaluated. Images were classified as 14 tissue type and trained. The accuracy for classification of each tissue type was 86.9%. Images were classified as either benign or malignant. The accuracy, sensitivity and specificity were 95.4%, 93.2% and 96.4%, respectively. We achieved 95% of accuracy for classification from ultrasound imaging using CNN. Deep learning could potentially help detecting and diagnosing the breast cancer, improving accuracy and productivity of diagnosing breast cancer by physician.

Improvement of Lesion Detection by Complete Angular Compound Ultrasonic Elastography.

Quasi-static ultrasound elastography is an emerging diagnostic imaging modality for determining the stiffness of pathologically changed soft tissues, which do not show significant differences in acoustic impedance for B-mode imaging. Although some methods were applied to improve the signal-to-noise ratio (SNRe) and contrast-to-noise ratio (CNRe) of the constructed elastogram, nonuniform strain distribution at the internal boundary of a hard inclusion, even with the uniform displacement on the surface, is an inherent mechanical effect and results in distortion at the detected lesion boundary. To overcome such stress concentrations, a new elastographic modality was proposed, where the elastograms from different angles throughout 360° were compounded. The strain field and subsequent ultrasound images were calculated using the finite element method (FEM) and Field II, respectively, from which the elastograms were constructed. The performance of complete angular compound elastography with varied interval angles, lesion sizes, and ratios of Young's moduli of the lesion to the background was simulated and compared with that of conventional axial strain elastography. It is found that viewing the lesion from only about 10 angles (interval of 36°) would significantly improve the image quality of elastogram (increasing SNRe by at least 13% and CNRe by at least 5.8 dB), reduce the lesion distortion in the lateral direction, and enhance the sensitivity, resolution, and accuracy of lesion detection. A preliminary phantom study showed similar improvements. Altogether, complete angular compound elastography improves the elastogram quality and reduces the mechanical effects in lesion detection.

MRI Contrast Agents: Evolution of Clinical Practice and Dose Optimization.

Accurate detection of lesions throughout the body is of paramount importance in contrast-enhanced magnetic resonance imaging (MRI). Optimal contrast agent performance is therefore of great importance and given the number of MRI contrast agent options today, this topic is of much ongoing study. The goal of this review article is to bring the read up to date on pertinent articles that relate to the evolution of radiological clinical practice and dose optimization pertaining to gadolinium contrast agents.

Accuracy of pre-contrast imaging in abdominal magnetic resonance imaging of pediatric oncology patients.

Safety concerns are increasingly raised regarding the use of gadolinium-based contrast media for MR imaging. To determine the accuracy of pre-contrast abdominal MR imaging for lesion detection and characterization in pediatric oncology patients. We included 120 children (37 boys and 83 girls; mean age 8.94years) referred by oncology services. Twenty-five had MRI for the first time and 95 were follow-up scans. Two authors independently reviewed pre-contrast MR images to note the following information about the lesions: location, number, solid vs. cystic and likely nature. Pre- and post-contrast imaging reviewed together served as the reference standard. The overall sensitivity was 88% for the first reader and 90% for the second; specificity was 94% and 91%; positive predictive value was 96% and 94%; negative predictive value was 82% and 84%; accuracy of pre-contrast imaging for lesion detection as compared to the reference standard was 90% for both readers. The difference between mean number of lesions detected on pre-contrast imaging and reference standard was not significant for either reader (reader 1, P = 0.072; reader 2, P = 0.071). There was substantial agreement (kappa values of 0.76 and 0.72 for readers 1 and 2) between pre-contrast imaging and reference standard for determining solid vs. cystic lesion and likely nature of the lesion. The addition of post-contrast imaging increased confidence of both readers significantly (P < 0.0001), but the interobserver agreement for the change in confidence was poor (kappa 0.12). Pre-contrast abdominal MR imaging has high accuracy in lesion detection in pediatric oncology patients and shows substantial agreement with the reference standard for characterization of lesions. Gadolinium-based contrast media administration cannot be completely eliminated but can be avoided in many cases, with the decision made on a case-by-case basis, taking into consideration location and type of tumor.

Diffusion-weighted imaging in the assessment of prostate cancer: Comparison of zoomed imaging and conventional technique

PurposeTo compare reduced field-of-view (zoomed) diffusion-weighted imaging (DWI) and conventional DWI in the evaluation of prostate cancer with respect to lesion detection, image quality and alignment accuracy. Material and methodsThe study was carried out in accordance with the Declaration of Helsinki and was approved by the institutional review board. Image data of 29 histology-proven prostate cancer lesions in 15 patients were evaluated. All patients underwent both conventional DWI and zoomed DWI at 3T. Zoomed DWI and conventional DWI sequences were analysed qualitatively and quantitatively. Subjective image quality, visual distortion and presence of artefacts were rated on a 5-point Likert scale (1=excellent) by two readers in consensus. Lesion conspicuity, sensitivity and specificity in lesion detection were evaluated and compared for both DWI sequences using ROC curves and area under the curve (AUC). To analyze the geographic distortion in DWI the alignment accuracy of prostate and lesions was measured in three spatial dimensions referring to the T2-weighted anatomical images as reference. In a region of interest (ROI) evaluation, ADC values were measured in prostate tissue and malignant lesions. Comparison of qualitative and quantitative parameters was performed using Wilcoxon test with subsequent Bonferroni correction. ResultsSubjective image quality was rated significantly higher in zoomed DWI compared to conventional DWI (2.1±0.9 vs. 2.7±0.9; p=0.0375). Visual distortion and artefacts were reduced in zoomed DWI without reaching statistical significance (1.8±0.7 vs. 2.4±1.0 and 2.1±1.0 vs. 2.5±1.0). Sensitivity and specificity of zoomed and conventional DWI were not significantly different. Zoomed DWI had a slightly higher AUC compared to conventional DWI without significant difference (0.82 versus 0.78; p=0.0576). Lesion conspicuity did not significantly differ between zoomed DWI and conventional DWI (1.8±0.8 vs. 1.9±1.0; p=0.8523). The alignment accuracy of zoomed DWI was significantly higher regarding both the prostate gland and lesions (deviation of outer contours of lesions in sagittal plane: 3±4mm vs. 5±3mm; p=0.0008). ADC tended to be higher in zoomed DWI without statistical significance (ADCmean in peripheral zone: 1.7±0.2×10−3mm2/s vs. 1.5±0.4×10−3mm2/s; ADCmean in lesion: 1.0±0.71×10−3mm2/s vs. 0.8±0.2×10−3mm2/s). ConclusionsZoomed technique offers improved image quality for diffusion-weighted imaging of the prostate with reduced image distortion both for the whole gland as well as for cancer lesions and at least comparable diagnostic performance. The zoomed technique could be useful for multiparametric tissue characterization but also for biopsy and radiation therapy planning.

MDCT imaging of traumatic brain injury.

The aim of emergency imaging is to detect treatable lesions before secondary neurological damage occurs. CT plays a primary role in the acute setting of head trauma, allowing accurate detection of lesions requiring immediate neurosurgical treatment. CT is also accurate in detecting secondary injuries and is therefore essential in follow-up. This review discusses the main characteristics of primary and secondary brain injuries.

Automatic Detection of Pectoral Muscle Region for Computer-Aided Diagnosis Using MIAS Mammograms.

The computer-aided detection (CAD) systems have been developed to help radiologists with the early detection of breast cancer. This system provides objective and accurate information to reduce the misdiagnosis of the disease. In mammography, the pectoral muscle region is used as an index to compare the symmetry between the left and right images in the mediolateral oblique (MLO) view. The pectoral muscle segmentation is necessary for the detection of microcalcification or mass because the pectoral muscle has a similar pixel intensity as that of lesions, which affects the results of automatic detection. In this study, the mammographic image analysis society database (MIAS, 322 cases) was used for detecting the pectoral muscle segmentation. The pectoral muscle was detected by using the morphological method and the random sample consensus (RANSAC) algorithm. We evaluated the detected pectoral muscle region and compared the manual segmentation with the automatic segmentation. The results showed 92.2% accuracy. We expect that the proposed method improves the detection accuracy of breast cancer lesions using a CAD system.

Evaluation and reduction of respiratory motion artifacts in small animal SPECT with GATE

The degradation of image quality caused by respiration is a major impediment to accurate lesion detection in single photon emission computed tomography (SPECT) imaging. This study was performed to evaluate the effects of lung motion on image quantification. A small animal SPECT system with NaI(Tl) was modeled in the Geant4 application for tomographic emission (GATE) simulation for a lung lesion using a 4D mouse whole-body phantom. SPECT images were obtained using 120 projection views acquired from 0o to 360o with a 3o step. Slices were reconstructed using ordered subsets expectation maximization (OS-EM) without attenuation correction with five iterations and four subsets. Image quality was compared between the static mode without respiratory motion, and dynamic mode with respiratory motion in terms of spatial resolution was measured by the full width at half maximum (FWHM), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). The FWHM of the non-gated image and the respiratory gated image were also compared. Spatial resolution improved as activity increased and lesion diameter decreased in the static and dynamic modes. The SNR and CNR increased significantly as lesion activity increased and lesion diameter decreased. Our results show that respiratory motion leads to reduced contrast and quantitative accuracy and that image quantification depends on both the amplitude and the pattern of the respiratory motion. We verified that respiratory motion can have a major effect on the accuracy of measurement of lung lesions and that respiratory gating can reduce activity smearing on SPECT images.

Evaluation and reduction of respiratory motion artifacts in small animal SPECT with GATE

The degradation of image quality caused by respiration is a major impediment to accurate lesion detection in single photon emission computed tomography (SPECT) imaging. This study was performed to evaluate the effects of lung motion on image quantification. A small animal SPECT system with NaI(Tl) was modeled in the Geant4 application for tomographic emission (GATE) simulation for a lung lesion using a 4D mouse whole-body phantom. SPECT images were obtained using 120 projection views acquired from 0̂ to 360̂ with a 3̂ step. Slices were reconstructed using ordered subsets expectation maximization (OS-EM) without attenuation correction with five iterations and four subsets. Image quality was compared between the static mode without respiratory motion, and dynamic mode with respiratory motion in terms of spatial resolution was measured by the full width at half maximum (FWHM), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). The FWHM of the non-gated image and the respiratory gated image were also compared. Spatial resolution improved as activity increased and lesion diameter decreased in the static and dynamic modes. The SNR and CNR increased significantly as lesion activity increased and lesion diameter decreased. Our results show that respiratory motion leads to reduced contrast and quantitative accuracy and that image quantification depends on both the amplitude and the pattern of the respiratory motion. We verified that respiratory motion can have a major effect on the accuracy of measurement of lung lesions and that respiratory gating can reduce activity smearing on SPECT images.

A hierarchical model for automated breast lesion detection from ultrasound 3D data.

Ultrasound imaging plays an important role in breast cancer screening for which early and accurate lesion detection is crucial for clinical practice. Many researches were performed on supporting the breast lesion detection based on ultrasound data. In the paper, a novel hierarchical model is proposed to automatically detect breast lesion from ultrasound 3D data. The model simultaneously considers the data information from low-level to high-level for the detection by processing with a joint probability. For each layer of the model, the corresponding algorithm is performed to denote the certain level image information. A dynamic programming approach is applied to efficiently obtain the optimal solution. With a preliminary dataset, the superior performance of the proposed model has been demonstrated for the automated detection of breast lesion with 0.375 false positive per case at 91.7% sensitivity.

Tu1959 Potential Role of EZH2 Expression in Pancreatic Adenocarcinoma Cells

G A A b st ra ct s screening via CT or MRI with EUS. Follow-up was recommended at 12 month intervals unless an abnormality warranted repeat screening at an earlier interval. All patients who presented for initial screening evaluation but failed to proceed to imaging evaluation tests were excluded. Results: There were 107 in the final cohort [Median Age 53 (29-84); 73 F (68.2%)] with follow-up of 0-109 months. There were 9 BRCA1 (3 with ≥1 first degree relative PC family history (FDR), 3 ≥1 second degree PC family history (SDR) and 3 with no PC family history). There were 27 BRCA2 (7 with ≥1 FDR, 3 with ≥1 SDR, 15 with no family history of PC); 2 HNPCC (1 ≥1 FDR and 1 ≥1 SDR), 4 Peutz-Jeghers with no PC family history, 26 Familial PC (≥2 FDR), 57 with PC family history not meeting Familial PC requirements. There were 19 (17.8%) initially screened via CT/MRI/EUS, 41 (38.3%) with CT/EUS, 5 (4.7%) with MRI/EUS, 27 (25.2%) with CT only , 5 (4.7%) with MRI only, and 8 (7.5%) with EUS only. The overall lesion detection accuracy for CT, MRI, and EUS were 20.0%, 18.8%, 34.2%. EUS was more accurate than MRI or CT at detecting cystic masses (EUS 23.3%, MRI 10.0%, CT 10.0%), side branch dilations (EUS 4.1%, MRI 0.0%, CT 0.0% ), parenchymal abnormalities (EUS 15.1%,MRI 6.3%, CT 1.1% ), and PD abnormalities (EUS 4.1%, MRI 0.0%, CT 1.1% ). Patients that had positive screening had a higher median age (57 vs 49), smoking history (30.6 vs 16.9%), and alcohol usage (50 vs 35.2%) than patients without abnormalities. Among HRI subgroups, more patients with initial pancreatic abnormalities on screening came from the Familial PC group (38.9 vs 16.9%). On the first follow-up imaging examination, EUS had greater accuracy at detecting all lesion types compared to MRI and CT (EUS 31.8%, MRI 8.0%, CT 4.3%), and all lesion sub-types that included cystic masses (EUS 18.2%, MRI 8.0%, CT 4.3%), side branch dilations (EUS 10.4%, MRI 0%, CT 0%), PD abnormalities (EUS 10.4%, MRI 0%, CT 0%), and other lesions (EUS 10.4%, MRI 0%, CT 0%). With regards to HRI characteristics, they also tended to be from the Familial PC HRI subgroup than other HRI groups (36.3% vs 28.8%). Conclusion: EUS appears to have a higher diagnostic yield compared to CT and MRI at detecting cystic masses, side branch dilations, parenchymal and PD abnormalities on initial screening and all pancreatic abnormalities on follow-up.

Europium-engineered iron oxide nanocubes with high T1 and T2 contrast abilities for MRI in living subjects.

Magnetic resonance imaging (MRI) contrast agents with both positive (T1) and negative (T2) contrast abilities are needed in clinical diagnosis for fault-free accurate detection of lesions. We report a facile synthesis of europium-engineered iron oxide (EuIO) nanocubes as T1 and T2 contrast agents for MRI in living subjects. The Eu(iii) oxide-embedded iron oxide nanoparticles significantly increase the T1 relaxivity with an enhanced positive contrast effect. EuIO nanocubes with 14 nm in diameter showed a high r1 value of 36.8 mM(-1) s(-1) with respect to total metal ions (Fe + Eu), which is about 3 times higher than that of Fe3O4 nanoparticles with similar size. Moreover, both r1 and r2 values of EuIO nanocubes can be tuned by varying their sizes and Eu doping ratios. After citrate coating, EuIO nanocubes can provide enhanced T1 and T2 contrast effects in small animals, particularly in the cardiac and liver regions. This work may provide an insightful strategy to design MRI contrast agents with both positive and negative contrast abilities for biomedical applications.

Block-Based Statistics for Robust Non-parametric Morphometry.

Automated algorithms designed for comparison of medical images are generally dependent on a sufficiently large dataset and highly accurate registration as they implicitly assume that the comparison is being made across a set of images with locally matching structures. However, very often sample size is limited and registration methods are not perfect and may be prone to errors due to noise, artifacts, and complex variations of brain topology. In this paper, we propose a novel statistical group comparison algorithm, called block-based statistics (BBS), which reformulates the conventional comparison framework from a non-local means perspective in order to learn what the statistics would have been, given perfect correspondence. Through this formulation, BBS (1) explicitly considers image registration errors to reduce reliance on high-quality registrations, (2) increases the number of samples for statistical estimation by collapsing measurements from similar signal distributions, and (3) diminishes the need for large image sets. BBS is based on permutation test and hence no assumption, such as Gaussianity, is imposed on the distribution. Experimental results indicate that BBS yields markedly improved lesion detection accuracy especially with limited sample size, is more robust to sample imbalance, and converges faster to results expected for large sample size.

Cone beam computed tomography and periapical lesions: a systematic review analysing studies on diagnostic efficacy by a hierarchical model.

To evaluate using a systematic review approach the diagnostic efficacy of CBCT for periapical lesions, focusing on the evidence level of the included studies using a six-tiered hierarchical model. The MEDLINE bibliographic database was searched from 2000 to July 2013 for studies evaluating the potential of CBCT imaging in the diagnosis and planning of treatment for periapical lesions. The search strategy was limited to English language publications using the following combined terms in the search strategy: apical pathology or endodontic pathology or periapical or lesion or healing and CBCT or cone beam CT. The diagnostic efficacy level of the studies was assessed independently by four reviewers. The search identified 25 publications that qualitatively or quantitatively assessed the use of CBCT for the diagnosis of periapical lesions, in which the methodology/results comprised at least one of the following parameters: the methods, the imaging protocols or qualitative/quantitative information on how CBCT influenced the diagnosis and/or treatment plan. From the assessed studies, it can be concluded that although there is a tendency for a higher accuracy for periapical lesion detection using CBCT compared to two-dimensional imaging methods, no studies have been conducted that justify the standard use of CBCT in diagnosing periapical lesions. In addition, it should be considered that, at the present time, the efficacy of CBCT as the diagnostic imaging method for periapical lesions has been assessed merely at low diagnostic efficacy levels.

Initial experiences with MR Image-guided laparoscopic microwave coagulation therapy for hepatic tumors.

Percutaneous thermal ablation is used for treating hepatic tumors. Recent advances in laparoscopy and imaging modalities have led to the development of a novel image-guided minimally invasive loco-regional treatment. The aim of this trial was to apply laparoscopic assistance to magnetic resonance (MR) image-guided thermoablation instead of ultrasonography, because of its various advantages. Patients with hepatic tumors and liver cirrhosis underwent magnetic resonance (MR) image-guided laparoscopic microwave coagulation therapy using a borescope and endoscopic forceps. Six cases of laparoscopic microwave coagulation treatment using MR image guidance were successfully performed between January 2000 and December 2008. Tumors were detected, punctured, and ablated in an open-configured MR scanner. A total of nine hepatocellular carcinoma nodules were preoperatively identified in S3, S5 and S6 (mean diameter=20.8±5.4mm). MR-guided microwave coagulation was laparoscopically achieved in all patients without any significant complications that required invasive treatment. The mean length of the operation was 275.3±60.5min, and the mean postsurgical hospital stay was 10.0±2.3days. Postoperative confirmation scanning was performed without moving the patients. MR-guided laparoscopic microwave coagulation therapy is an effective treatment for hepatic tumors adjacent to other organs, as it allows for more accurate detection of lesions and for tumors to be treated safely while avoiding adjacent organs. It is less invasive than conventional procedures, because the MR real-time guidance enabled continuous monitoring throughout the procedure.

Labral and cartilage abnormalities in young patients with hip pain: accuracy of 3-Tesla indirect MR arthrography.

Assess the diagnostic accuracy of 3-T indirect magnetic resonance arthrography (iMRA) for hip cartilage and labral pathology detection using arthroscopy as the reference standard and compare it to the published performance of direct magnetic resonance arthrography (dMRA). Between 2009 and 2011, 290 patients suspected of having femoroacetabular impingement underwent iMRA. Our study group consisted of 41 of these patients (17 males, mean age 35years; 24 females, mean age 33years) who did not have a prior history of hip surgery and who subsequently underwent arthroscopy. Two experienced musculoskeletal radiologists separately evaluated the randomized and anonymized studies for the presence and quadrant location of labral and cartilage pathology. These recorded data were compared to arthroscopic reports. Forty-one patients had labral pathology, 34 patients had acetabular and 5 patients had femoral cartilage pathology at arthroscopy. Sensitivity, specificity, accuracy, negative- and positive-predictive values for labral lesion detection were respectively 98, 99, 99, 99 and 98%; for acetabular cartilage lesion detection they were 69, 98, 89, 87 and 95%; for femoral cartilage lesion detection they were 69, 95, 93 and 39%. Sensitivities of iMRA by quadrant (anteroinferior, anterosuperior, posteroinferior, posterosuperior) for the labrum were 100.0, 95.0, NA and 85.7%, for acetabular cartilage were NA, 58.8, NA and 39.5% and for femoral cartilage were 50.0, 33.3, 75.0 and 75.0%). NA indicates results not available because of the absence of findings in those quadrants. Specificities of iMRA by quadrant (anteroinferior, anterosuperior, posteroinferior, posterosuperior) for the labrum (95.0, 100.0, 95.1, 67.5%), acetabular (100.0, 85.7, 92.6, 79.5%) and femoral cartilage (100.0, 94.7, 96.2, 85.9%). iMRA at 3T is accurate in detecting labral pathology suggesting that it is a viable alternative to dMRA.

Can single-phase dual-energy CT reliably identify adrenal adenomas?

To evaluate whether single-phase dual-energy-CT-based attenuation measurements can reliably differentiate lipid-rich adrenal adenomas from malignant adrenal lesions. We retrospectively identified 51 patients with adrenal masses who had undergone contrast-enhanced dual-energy-CT (140/100 or 140/80 kVp). Virtual non-contrast and colour-coded iodine images were generated, allowing for measurement of pre- and post-contrast density on a single-phase acquisition. Adrenal adenoma was diagnosed if density on virtual non-contrast images was ≤10 HU. Clinical follow-up, true non-contrast CT, PET/CT, in- and opposed-phase MRI, and histopathology served as the standard of reference. Based on the standard of reference, 46/57 (80.7%) adrenal masses were characterised as adenomas or other benign lesions; 9 malignant lesions were detected. Based on a cutoff value of 10 HU, virtual non-contrast images allowed for correct identification of adrenal adenomas in 33 of 46 (71%), whereas 13/46 (28%) adrenal adenomas were lipid poor with a density ≥10 HU. Based on the threshold of 10 HU on the virtual non-contrast images, the sensitivity, specificity, and accuracy for detection of benign adrenal lesions was 73%, 100%, and 81% respectively. Virtual non-contrast images derived from dual-energy-CT allow for accurate characterisation of lipid-rich adrenal adenomas and can help to avoid additional follow-up imaging. • Adrenal adenomas are a common lesion of the adrenal glands. • Differentiation of benign adrenal adenomas from malignant adrenal lesions is important. • Dual-energy based virtual non-contrast images help to evaluate patients with adrenal adenomas.

Diagnostic accuracy of noncontrast MRI for detection of glenohumeral cartilage lesions: a prospective comparison to arthroscopy

The purpose of this study was to determine the prevalence of glenohumeral articular cartilage lesions in patients with rotator cuff tendinopathy and to assess the accuracy of noncontrast magnetic resonance imaging (MRI) in detecting these defects compared with the "gold standard" of arthroscopy. Noncontrast MRI studies obtained in 84 consecutive patients undergoing shoulder arthroscopy for rotator cuff tendinopathy (mean age, 54.8 years; range, 17-82 years) were prospectively evaluated for glenohumeral cartilage lesions. Two fellowship-trained, experienced musculoskeletal radiologists were blinded from the arthroscopic findings and independently evaluated the glenoid and humeral head cartilage on 2 separate occasions. At arthroscopy, cartilage lesions of the humeral head were detected in 23 patients (frequency, 27.4%), and glenoid cartilage lesions were found in 20 patients (frequency, 23.8%). For detection of a humeral lesion on MRI, the radiologists' combined accuracy was 78%, sensitivity was 43%, and specificity was 91%. The combined accuracy for detection of glenoid lesions on MRI was 84%, sensitivity was 53%, and specificity was 93%. Combining the readers, low-grade lesions (International Cartilage Repair Society grades 1 and 2) of the glenoid and humerus were read as negative on MRI in 63% and 86% of cases, respectively. Overall accuracy of noncontrast MRI for detection of glenohumeral articular cartilage lesions is good; however, interpretation is reader dependent, and accuracy is significantly reduced for detection of low-grade lesions. On the basis of these findings, we recommend that patients with rotator cuff tendinopathy undergoing arthroscopy be informed that the presence and severity of cartilage lesions may be underestimated on MRI.

Digital breast tomosynthesis and digital mammography: A comparison of figures of merit for various average glandular doses

Previous studies on the application of tomosynthesis to breast imaging have demonstrated the potential of digital breast tomosynthesis (DBT). DBT can improve the specificity of digital mammography (DM) through improved marginal visibility of lesions and early breast cancer detection for women with dense breasts. To investigate possible improvements in the accuracy of lesion detection with DBT systems as compared to DM, we conducted a quantitative evaluation by using simulated lesions embedded in a breast phantom. A prototype DBT and dedicated DM system were used in this study. For the DBT system, the average glandular dose (AGD) was calculated using a formalism that was a simple extension of mammography dosimetry. The DBT and the DM images were acquired with average glandular doses (AGDs) ranging from 1 to 4 mGy. To analyze the results objectively, we calculated metrics for in-plane lesion visibility in the form of the contrast-to-noise ratio for the in-focus plane from the DBT reconstruction image and from the craniocaudal (CC) image from the DM system. The imaging performance of DBT was quantitatively compared with that of DM in terms of the figure of merit. Although the DM showed better results in terms of the contrast-to-noise ratio (CNR) of the mass due to the reduced overlapping of tissue and lesion, an increase in breast thickness of over 3 cm increased the CNR of the mass with the DBT system. For microcalcification detection, the DBT system showed significantly higher CNR than the DM system and gave better predictions of the microcalcification size. We compared the performances of the DM and the DBT systems for various AGDs and breast thicknesses. In conclusion, the results indicate that the DBT systems can play an important role in the detection of masses or microcalcifications without severe compression.

Effect of Reconstruction Parameters in High-Definition PET/CT on Assessment of Lymph Node Metastases in Head and Neck Squamous Cell Carcinoma

Therapy outcome and overall survival in patients with head and neck squamous cell carcinoma (HNSCC) is influenced by precise localization of the primary tumor and detection of lymph node metastasis involvement at the time of initial diagnosis. Only accurate preoperative staging can improve primary tumor response and avoid early locoregional recurrence with lymph node metastases. The purpose of this study was the optimization of reconstruction parameters in high-definition PET/CT for the improved diagnostic assessment of lymph node metastases. In the experimental study, image contrast and signal-to-noise ratio were evaluated using a Jaszczak phantom. In the clinical study, 54 patients underwent head and neck imaging on a PET/CT scanner. Diagnostic findings were correlated with postoperative histopathology. For the 54 patients, 123 lymph nodes were evaluated on PET and histologically correlated with the neck dissection specimen. Forty-one lymph nodes were benign, and 82 findings were confirmed as being malignant. Both experimental and clinical studies were reconstructed into a 200 × 200 matrix using a 3-dimensional iterative reconstruction algorithm (ordered-subset expectation maximization [OSEM], 3 iterative steps, 24 subsets). Postfiltering with a 3-dimensional gaussian filter was applied. To study the effect of smoothing filter strength on the diagnostic accuracy of lymph node metastasis detection, 3 different cutoffs-1, 3, and 6 mm in full width at half maximum-were used to perform reconstructions. Phantom studies showed that images reconstructed with 3-mm gaussian postfiltering gained a higher image quality and signal-to-noise ratio. Overall sensitivities for correctly diagnosed lymph nodes were best in 3-mm postfiltered images. Best results for true-positive lymph node findings were achieved with 3-mm postfiltering. With 1-mm postfiltering, accurate lesion detection was not improved, because increasing sensitivity (95% true-positive) correlated with decreasing specificity (12% false-positive). For lymph node assessment on a high-resolution PET/CT scanner, we consider the OSEM algorithm with 3 iterations and 24 subsets, combined with 3-dimensional 3-mm gaussian postfiltering, to be optimal. The continuous application of presently established PET protocols in patients with HNSCC will prove whether current acquisition and reconstruction methods are valuable and should be maintained.