Magda Marcon

Age- and Level-Dependence of Fatty Infiltration in Lumbar Paravertebral Muscles of Healthy Volunteers.

BACKGROUND AND PURPOSE: Normative age-related decline in paravertebral muscle quality is important for reference to disease and risk identification in patients. We aimed to establish age- and vertebral level–dependence of paravertebral (multifidus and erector spinae) muscle volume and fat content in healthy adult volunteers. MATERIALS AND METHODS: In this prospective study multifidus and erector spinae fat signal fraction and volume at lumbar levels L1–L5 were measured in 80 healthy volunteers (10 women and men per decade, 20–62 years of age) by 2-point Dixon 3T MR imaging. ANOVA with post hoc Bonferroni correction compared fat signal fraction and volume among subgroups. Pearson and Spearman analysis were used for correlations (P < .05). RESULTS: Fat signal fraction was higher in women (17.8% ± 10.7%) than men (14.7% ± 7.8%; P < .001) and increased with age. Multifidus and erector spinae volume was lower in women (565.4 ± 83.8 cm3) than in men (811.6 ± 98.9 cm3; P < .001) and was age-independent. No differences in fat signal fraction were shown between the right and left paravertebral muscles or among the L1, L2, and L3 lumbar levels. The fat signal fraction was highest at L5 (women, 31.9% ± 9.3%; men, 25.7% ± 8.0%; P < .001). The fat signal fraction at L4 correlated best with total lumbar fat signal fraction (women, r = 0.95; men, r = 0.92, P < .001). Total fat signal fraction was higher in the multifidus compared with erector spinae muscles at L1–L4 for both sexes (P < .001). CONCLUSIONS: Lumbar paravertebral muscle fat content increases with aging, independent of volume, in healthy volunteers 20–62 years of age. Women, low lumbar levels, and the multifidus muscle are most affected. Further study examining younger and older subjects and the functional impact of fatty infiltrated paravertebral muscles are warranted.

Deep Learning in Mammography: Diagnostic Accuracy of a Multipurpose Image Analysis Software in the Detection of Breast Cancer

Objectives The aim of this study was to evaluate the diagnostic accuracy of a multipurpose image analysis software based on deep learning with artificial neural networks for the detection of breast cancer in an independent, dual-center mammography data set. Materials and Methods In this retrospective, Health Insurance Portability and Accountability Act-compliant study, all patients undergoing mammography in 2012 at our institution were reviewed (n = 3228). All of their prior and follow-up mammographies from a time span of 7 years (2008–2015) were considered as a reference for clinical diagnosis. After applying exclusion criteria (missing reference standard, prior procedures or therapies), patients with the first diagnosis of a malignoma or borderline lesion were selected (n = 143). Histology or clinical long-term follow-up served as reference standard. In a first step, a breast density-and age-matched control cohort was selected (n = 143) from the remaining patients with more than 2 years follow-up (n = 1003). The neural network was trained with this data set. From the publicly available Breast Cancer Digital Repository data set, patients with cancer and a matched control cohort were selected (n = 35 × 2). The performance of the trained neural network was also tested with this external data set. Three radiologists (3, 5, and 10 years of experience) evaluated the test data set. In a second step, the neural network was trained with all cases from January to September and tested with cases from October to December 2012 (screening-like cohort). The radiologists also evaluated this second test data set. The areas under the receiver operating characteristic curve between readers and the neural network were compared. A Bonferroni-corrected P value of less than 0.016 was considered statistically significant. Results Mean age of patients with lesion was 59.6 years (range, 35–88 years) and in controls, 59.1 years (35–83 years). Breast density distribution (A/B/C/D) was 21/59/42/21 and 22/60/41/20, respectively. Histologic diagnoses were invasive ductal carcinoma in 90, ductal in situ carcinoma in 13, invasive lobular carcinoma in 13, mucinous carcinoma in 3, and borderline lesion in 12 patients. In the first step, the area under the receiver operating characteristic curve of the trained neural network was 0.81 and comparable on the test cases 0.79 (P = 0.63). One of the radiologists showed almost equal performance (0.83, P = 0.17), whereas 2 were significantly better (0.91 and 0.94, P < 0.016). In the second step, performance of the neural network (0.82) was not significantly different from the human performance (0.77–0.87, P > 0.016); however, radiologists were consistently less sensitive and more specific than the neural network. Conclusions Current state-of-the-art artificial neural networks for general image analysis are able to detect cancer in mammographies with similar accuracy to radiologists, even in a screening-like cohort with low breast cancer prevalence.

Quantitative and qualitative MR‐imaging assessment of vastus medialis muscle volume loss in asymptomatic patients after anterior cruciate ligament reconstruction

Background: To quantitatively and qualitatively assess vastus medialis muscle atrophy in asymptomatic patients with anterior cruciate ligament reconstruction, using the nonoperated leg as control. Methods: Prospective Institutional Review Board approved study with written informed patient consent. Thirty‐three asymptomatic patients (men, 21; women,12) with ACL‐reconstruction underwent MR imaging of both legs (axial T1‐weighted spin‐echo and 3D spoiled dual gradient‐echo sequences). Muscle volume and average fat‐signal fraction (FSF) of the vastus medialis muscles were measured. Additionally, Goutallier classification was used to classify fatty muscle degeneration. Significant side differences were evaluated using the Wilcoxon test and, between volumes and FSF, using student t‐tests with P‐value < 0.05 and < 0.025, respectively. Results: The muscle volume was significantly smaller in the operated (mean ± SD, 430.6 ± 119.6 cm3; range, 197.3 to 641.7 cm3) than in the nonoperated leg (479.5 ± 124.8 cm3; 261.4 to 658.9 cm3) (P < 0.001). Corresponding FSF was 6.3 ± 1.5% (3.9 to 9.2%) and 5.8 ± 0.9% (4.0 to 7.4%), respectively, with a nonsignificant (P > 0.025) difference. The relative muscle‐volume and FSF differences were −10.1 ± 8.6% (7.1 to −30.1%) and 10.9 ± 29.4% (39.7 to 40.1%). The qualitative assessment revealed no significant differences (P > 0.1). Conclusion: A significant muscle volume loss of the vastus medialis muscle does exist in asymptomatic patients with ACL‐reconstruction, but without fatty degeneration. J. Magn. Reson. Imaging 2015;42:515–525.

Whole-body adipose tissue and lean muscle volumes and their distribution across gender and age: MR-derived normative values in a normal-weight Swiss population

To determine age‐ and gender‐dependent whole‐body adipose tissue and muscle volumes in healthy Swiss volunteers in Dixon MRI in comparison with anthropometric and bioelectrical impedance (BIA) measurements.

Evaluation of Reproducibility of Diffusion Tensor Imaging in the Brachial Plexus at 3.0 T

Objective The aim of this study was to evaluate the reproducibility of 3 T magnetic resonance imaging diffusion tensor imaging (DTI) of the brachial plexus in healthy subjects. Methods Ten healthy volunteers were included, and morphological and DTI sequences of the nerve roots of the brachial plexus from C5 to T1 of both sides were repeatedly acquired on a 3 T magnetic resonance system (MAGNETOM Skyra; Siemens Healthcare, Erlangen, Germany). A prototype diffusion-weighted single-shot echo-planar imaging sequence-enabling slice-specific shim adjustments was performed with b-values of 0 and 800 s/mm2 in 30 gradient directions, resulting in an acquisition time of about 6 minutes each in axial orientation. Between scans, subjects were moved and repositioned in the scanner, coils were reinserted, and new localizers were acquired. Image analysis was performed using MITK Diffusion software toolkit. Two independent readers performed diffusion data postprocessing, and regions of interest (ROIs) were set on the proximal postganglionic trunk at each spinal level, bilaterally to obtain values for fractional anisotropy (FA) and mean diffusivity (MD). Interreader and intrareader agreement as well as test-retest reproducibility of DTI metrics were assessed. Results Intraclass correlation coefficients (ICCs) for interreader and intrareader agreement did not differ significantly between measurements for FA and MD. In particular, ICCs for interreader agreement of FA ranged from 0.741 to 0.961 and that of MD ranged from 0.802 to 0.998, and ICCs for intrareader agreement of FA ranged from 0.759 to 0.949 and that of MD ranged from 0.796 to 0.998. The test-retest reproducibility of DTI metrics showed an overall moderate to strong correlation (r > 0.707), with few minor exceptions, for both FA and MD values. Conclusions Diffusion tensor imaging metrics in the brachial plexus are reproducible. Future applications of DTI for a possible clinical use should be further investigated.

Age- and Gender Dependent Liver Fat Content in a Healthy Normal BMI Population as Quantified by Fat-Water Separating DIXON MR Imaging.

Objectives To establish age- and sex-dependent values of magnetic resonance (MR) liver fat-signal fraction (FSF) in healthy volunteers with normal body-mass index (BMI). Methods 2-point mDIXON sequences (repetition time/echo time, 4.2msec/1.2msec, 3.1msec) at 3.0 Tesla MR were acquired in 80 healthy volunteers with normal BMI (18.2 to 25.7 kg/m2) between 20 and 62 years (10 men/10 women per decade). FSF was measured in 5 liver segments (segment II, III, VI, VII, VIII) based on mean signal intensities in regions of interest placed on mDIXON-based water and fat images. Multivariate general linear models were used to test for significant differences between BMI-corrected FSF among age subgroups. Pearson and Spearman correlations between FSF and several body measures were calculated. Results Mean FSF (%) ± standard deviations significantly differed between women (3.91 ± 1.10) and men (4.69 ± 1.38) and varied with age for women/men (p-value: 0.002/0.027): 3.05 ± 0.49/3.74 ± 0.60 (age group 20–29), 3.75 ± 0.66/4.99 ± 1.30 (30–39), 4.76 ± 1.16/5.25 ± 1.97 (40–49) and 4.09 ± 1.26/4.79 ± 0.93 (50–62). FSF differences among age subgroups were significant for women only (p = 0.003). Conclusions MR-based liver fat content is higher in men and peaks in the fifth decade for both genders.

Classification of breast cancer from ultrasound imaging using a generic deep learning analysis software: a pilot study

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 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.

First ultrasound diagnosis of BI-RADS 3 lesions in young patients: Can 6-months follow-up be sufficient to assess stability?

OBJECTIVES To evaluate the outcome of repeated short-term follow-up with ultrasound in no high-risk young patients with a BI-RADS3 lesion at first examination. METHODS In this IRB-approved study 492 women, aged 18-34 years (mean±standard deviation, 28±4.5years) with first breast ultrasound examination in 2012-2014 were retrospectively evaluated. Inclusion criteria were: at least one BI-RADS3 lesion and (a) biopsy/surgical excision or (b) follow-up of at least 18 months (including a 6-month follow-up). BI-RADS category assigned during follow-up and pathologic findings in cases undergoing biopsy/surgical excision were collected. At the 6- and 18-month follow-up the recommended biopsy rates (RBR) and the corresponding positive predictive value (PPV) were calculated. RESULTS In 97 patients, 151 BI-RADS3 lesions were identified. Biopsy/surgical excision was initially performed in 25/151 (16.5%) lesions. After 6-month, category was downgraded to BI-RADS1/2 in 23/126 (15.3%) and upgraded to BI-RADS4 in 9/126 lesions (7.1%). Pathological diagnosis of these lesions was fibroadenoma in 5 and benign phyllodes tumor in 4 cases (RBR 7%, PPVbio 44.4%). After 18-month one lesion was classified BI-RADS4 and pathological diagnosis was fibroadenoma (RBR 1.1%, PPVbio 0%). CONCLUSIONS Our preliminary data show that follow-up imaging performed after 18 months from a first BI-RADS3 diagnosis does not affect clinical treatment and 6-month follow-up may be sufficient to assess the stability of probably benign lesions.

Influence of breast cancer opportunistic screening on aesthetic surgical outcome: A single-center retrospective study in Switzerland

The aim of this retrospective study is to compare the surgical aesthetic outcome and breast cancer (BC) characteristics in patients with BC detected either by opportunistic screening or clinical diagnosis. 262 women undergoing surgery for BC between 2009 and 2012 were included. The following features were compared in the two groups of patients: (1) age at diagnosis; (2) family history of BC; (3) histology type; (4) tumor diameter; (5) local staging, and (6) type of surgical treatment. In 92/262 (35.1%) cases BC was detected by screening and 170/262 (64.9%) had clinical diagnosis. A positive family history and ductal carcinoma in situ diagnosis were more frequent in patients with clinical diagnosis (P = .001 and P < .0001 respectively). Mean maximum diameter of invasive cancers was significantly greater in symptomatic patients (P < .001). Breast conserving surgery was performed in 76/92 (82.6%) patients with screening and 115/170 (67.6%) with clinical diagnosis. Mastectomy was performed in 16/92 (17.4%) patients with screening and 55/170 (32.3%) with clinical diagnosis. Mastectomy was more frequent in patients with clinical diagnosis of BC (P = .010). No significant group differences were found regarding the other features. This study demonstrated that in opportunistic screening, breast conserving surgery may be applied in a higher number of cases compared to patients presenting with clinical diagnosis, thereby improving life quality of these patients.

A new method to reduce false positive results in breast MRI by evaluation of multiple spectral regions in proton MR-spectroscopy

OBJECTIVES To investigate whether the evaluation of multiple spectral regions can increase the diagnostic performance of 1H-MRS of the breast and reduce false positive findings. METHODS 93 patients (mean age 56 years, range 23-79) undergoing breast MRI for routine clinical indications on a 1.5T scanner were eligible for this IRB-approved prospective study. Suspicious enhancing lesions ≥8mm underwent single-voxel point-resolved 1H-MRS (PRESS, TR=2000ms, TE=272ms). Histology showed 69 malignant and 24 benign lesions. The Signal-to-Noise ratio of choline, olefinic acids and the water-to-methylene ratio were measured. The area under the ROC-curve (AUC) was used to evaluate single measurements. Combined diagnostic accuracy was explored using a 10-fold cross-validated Chi-squared automatic interaction detection (CHAID) analysis. Inter-reader agreement was evaluated in a subset of patients. RESULTS A significant AUC for differentiation between benign and malignant lesions was identified for choline (0.733, P=0.001), olefinic acids (0.769, P=0.0001) and water-to-methylene ratio (0.704, P=0.003). All three variables were included in a classification algorithm using CHAID methodology. Using this classification, 70.8% (17/24) false positive diagnoses in benign lesions would have been avoided. Inter-reader agreement was almost perfect. CONCLUSIONS The combined evaluation of multiple spectral regions can increase the diagnostic performance of 1H-MRS and potentially reduce false positive findings.