Search

In DCNNs, the number of parameters in pointwise convolutions rapidly grows due to the multiplication of the number of filters by the number of input channels that come from the previous layer. Our proposal makes pointwise convolutions parameter efficient via grouping filters into parallel branches or groups, where each branch processes a fraction of the input channels. However, by doing so, the learning capability of the DCNN is degraded. To avoid this effect, we suggest interleaving the output of filters from different branches at intermediate layers of consecutive pointwise convolutions. We applied our improvement to the EfficientNet, DenseNet-BC L100, MobileNet and MobileNet V3 Large architectures. We trained these architectures with the CIFAR-10, CIFAR-100, Cropped-PlantDoc and The Oxford-IIIT Pet datasets. When training from scratch, we obtained similar test accuracies to the original EfficientNet and MobileNet V3 Large architectures while saving up to 90% of the parameters and 63% of the flops.

The Food and Agriculture Organization (FAO) estimated that plant diseases cost the world economy $220 billion in 2019. In this paper, we propose a lightweight Deep Convolutional Neural Network (DCNN) for automatic and reliable plant leaf diseases classification. The proposed method starts by converting input images of plant leaves from RGB to CIE LAB coordinates. Then, L and AB channels go into separate branches along with the first three layers of a modified Inception V3 architecture. This approach saves from 1/3 to 1/2 of the parameters in the separated branches. It also provides better classification reliability when perturbing the original RGB images with several types of noise (salt and pepper, blurring, motion blurring and occlusions). These types of noise simulate common image variability found in the natural environment. We hypothesize that the filters in the AB branch provide better resistance to these types of variability due to their relatively low frequency in the image-space domain.

Background Several de-escalation approaches are under investigation in patients with HER2-positive, early-stage breast cancer. We assessed early metabolic responses to neoadjuvant trashizumab and pertuzumab using F-18-fluorodeoxyglucose (F-18-FDG)-PET (F-18-FDG-PET) and the possibility of chemotherapy de-escalation using a pathological response-adapted strategy. Methods We did a multicentre, randomised, open-label, non-comparative, phase 2 trial in 45 hospitals in Spain, France, Belgium, Germany, the UK, Italy, and Portugal. Eligible participants were women aged 18 years or older with centrally confirmed, HER2-positive, stage I-IIIA, invasive, operable breast cancer (1.5 cm tumour size) with at least one breast lesion evaluable by 18 F-FDG-PET, an Eastern Cooperative Oncology Group performance status of 0 or 1, and a baseline left ventricular ejection fraction of at least 55%. We randomly assigned participants (1:4), via an interactive response system using central block randomisation with block sizes of five, stratified by hormone receptor status, to either docetaxel (75 mg/m(2) intravenous), carboplatin (area under the concentration-time curve 6 mg/mL per min intravenous), trastuzumab (subcutaneous 600 mg fixed dose), and pertuzumab (intravenous 840 mg loading dose, 420 mg maintenance doses; group A); or trastuzumab and pertuzumab (group B). Hormone receptor-positive patients allocated to group B were additionally given letrozole if postmenopausal (2.5 mg/day orally) or tamoxifen if premenopausal (20 mg/day orally). Centrally reviewed F-18-FDG-PET scans were done before randomisation and after two treatment cycles. Patients assigned to group A completed six cycles of treatment (every 3 weeks) regardless of F-18-FDG-PET results. All patients assigned to group B initially received two cycles of trastuzumab and pertuzumab. F-18-FDG-PET responders in group B continued this treatment for six further cycles; F-18-FDG-PET non-responders in this group were switched to six cycles of docetaxel, carboplatin, trastuzumab, and pertuzumab. Surgery was done 2-6 weeks after the last dose of study treatment. Adjuvant treatment was selected according to the neoadjuvant treatment administered, pathological response, hormone receptor status, and clinical stage at diagnosis. The coprimary endpoints were the proportion of F-18-FDG-PET responders in group B with a pathological complete response in the breast and axilla (ypT0/is ypN0) as determined by a local pathologist after surgery after eight cycles of treatment, and 3-year invasive disease-free survival of patients in group B, both assessed by intention to treat. The definitive assessment of pathological complete response was done at this primary analysis; follow-up to assess invasive disease-free survival is continuing, hence these data are not included in this Article. Safety was assessed in all participants who received at least one dose of study drug. Health-related quality-of-life was assessed with EORTC QLQ-C30 and QLQ-BR23 questionnaires at baseline, after two cycles of treatment, and before surgery. This trial is registered with EudraCT (2016-002676-27) and ClinicalTrials.gov (NCT03161353), and is ongoing. Findings Between rune 26, 2017, and April 24, 2019, we randomly assigned 71 patients to group A and 285 to group B. Median follow-up was 5.7 months (IQR 5.3-6.0). 227 (80%) of 285 patients in group B were F-18-FDGPET responders, of whom 86 (37.9%, 95% CI 31.6-44-5; p

Il potassio è il principale ione intracellulare e la sua omeostasi è finemente regolata dall’apparato renale e gastrointestinale. L’insufficienza renale e l’iperpotassiemia sono condizioni di comune riscontro nei pazienti con scompenso cardiaco, risultato di una complessa interazione tra cuore e rene (es. sindrome cardio-renale) e degli effetti collaterali dei farmaci comunemente impiegati per il trattamento della malattia cardiaca (es. inibitori del sistema renina-angiotensina-aldosterone). Sebbene l’iperpotassiemia incrementi il rischio di disturbi di conduzione ed aritmie minacciose per la vita, il suo significato prognostico nello scompenso cardiaco è incerto. L’iperpotassiemia e la progressione del danno renale costituiscono i principali limiti all’introduzione e alla titolazione delle terapie eziologiche dello scompenso cardiaco. L’ingresso in commercio di nuovi farmaci per la prevenzione e il trattamento cronico dell’iperpotassiemia permette di introdurre e modulare la terapia anti-neurormonale in pazienti con scompenso cardiaco altrimenti esclusi per i livelli di potassio sierico eccessivamente elevati. Questa rassegna tratta gli aspetti fisiopatologici ed epidemiologico-prognostici dell’iperpotassiemia, offrendo, inoltre, un’analisi sulle possibili strategie di trattamento della disionia nello scompenso cardiaco.

BACKGROUND/PURPOSE: The main indication of covered stents (CS) is coronary artery perforation (CAP), but, they have been increasingly used in other scenarios. Data on the long-term follow-up of CS is limited, and no studies have been conducted specifically using new-generation polyurethane-covered cobalt-chromium Papyrus CS. PURPOSE: to evaluate the clinical outcomes after hospital discharge of Papyrus CS and to compare their outcome after implantation in CAP or coronary artery aneurysms (CAA). METHODS/MATERIALS: We evaluated the baseline clinical characteristics, lesion subsets, procedural features and the outcomes after initial discharge of Papyrus CS implanted in 17 high-PCI-volume centers. RESULTS: 127 Papyrus CS were implanted in 108 patients (68?±?1?years; 82.8% male) admitted for stable coronary disease (32.3%), NSTEMI (42.4%) or STEMI (25.3%). The number of CS per patient was 1.2?±?0.6 (diameter: 3.5?±?1.7?mm; length: 18.5?±?3.7?mm). Angiographic success rate was 96%. CS diameter was larger in CAA (CAP:3.04?±?0.5?mm vs CAA:4.1?±?2.7?mm; p?=?.022). Intracoronary imaging techniques were used more frequently in CAA (p?.0001). After a mean follow-up of 22?±?16?months, the major cardiovascular adverse events (MACE) rate was 7.1% [cardiac death: 2%, Myocardial infarction: 5%, Target Lesion Revascularization: 5% and Stent Thrombosis (ST): 3%]. MACE rate was similar in CAP (7.7%) and CAA (7.1%) (p?=?.9). However, CAA showed a higher ST rate (CAP: 0% vs CA: 7.1%; p?=?.04). CONCLUSION: After hospital discharge, clinical outcomes after Papyrus CS implantation are acceptable (considering the clinical scenario and compared with other treatment alternatives) with no significant differences in the MACE rate between those implanted in CAA or in CAP. However, CAA group showed a higher ST rate.

Simple Summary Considering the favorable overall safety profile of trastuzumab emtansine (T-DM1), the low expected rate of cardiotoxicity, and the synergistic effect of anthracyclines with Human Epidermal Growth Factor Receptor 2 (HER2)-targeting agents, it is hypothesized that T-DM1 may be safely combined with non-pegylated liposomal doxorubicin (NPLD). In the THELMA trial, the effect of adding NPLD to T-DM1 was evaluated with the aim of enhancing T-DM1 efficacy using an extensive cardiological assessment in trastuzumab- and taxane-pretreated patients with HER2-positive metastatic breast cancer. Despite an unlikely drug synergism, this combination was generally well tolerated without clinically relevant worsening of cardiac function. No relationship was identified between early predictors of heart failure and left ventricular ejection fraction changes. Thus, the combination of T-DM1 plus NPLD is safe, but this regimen does not seem to improve T-DM1 antitumor activity in this setting. The paper assesses the dose-limiting toxicities and the maximum tolerated dose (MTD) of trastuzumab emtansine (T-DM1) combined with non-pegylated liposomal doxorubicin (NPLD) in HER2-positive (HER2+) metastatic breast cancer (MBC). This single-arm, open-label, phase Ib trial (NCT02562378) enrolled anthracycline-naive HER2+ MBC patients who had progressed on trastuzumab and taxanes. Patients received a maximum of 6 cycles of NPLD intravenously (IV) at various dose levels (45, 50, and 60 mg/m(2)) in the "3 plus 3" dose-escalation part. During expansion, they received 60 mg/m(2) of NPLD every 3 weeks (Q3W) plus standard doses of T-DM1. The MTD was T-DM1 3.6 mg/kg plus NPLD 60 mg/m(2) administered IV Q3W. No clinically relevant worsening of cardiac function was observed. Among all evaluable patients, the overall response rate was 40.0% (95%CI, 16.3-67.7) with a median duration of response of 6.9 months (95%CI, 4.8-9.1). Clinical benefit rate was 66.7% (95%CI, 38.4-88.2) and median progression-free survival was 7.2 months (95%CI, 4.5-9.6). No significant influence of NPLD on T-DM1 pharmacokinetics was observed. The addition of NPLD to T-DM1 is feasible but does not seem to improve the antitumor efficacy of T-DM1 in HER2+ MBC patients.

According to the new EU Regulation (2017/2158), which has established new “mitigation measures and benchmark levels for the reduction of acrylamide in foods”, there is the necessity to adapt different process methods in order to reduce the acrylamide formation in deep-fat fried potato-based products, and at the same time to maintain the quality of final products acceptable. Among strategies suggested for the reduction of acrylamide in potato crisps, pulsed electric fields (PEF) represent a promising innovative technology with the potentiality to remove Maillard reaction substrates, such as reducing sugars and free asparagine, in raw potato tissues. Recently, it has been reported that the strain L1 of the yeast Aureobasidium pullulans has the ability to assimilate asparagine in food systems. On the basis of this consideration, the aim of the study was to evaluate the possibility of the reduction of acrylamide content in potato crisps, pre-treating the raw material using a combination of PEF treatment to enhance the extractability of asparagine and a subsequent treatment with aqueous suspension of the aforementioned yeast. Potato tubers (Solanum tuberosum, var. Lady Claire) were manually peeled and sliced and submitted to PEF pre-treatment using a lab-scale unit delivering rectangular shape pulses. An electric field strength of 1.5 kV/cm and 1000 pulses were used. Consequently, PEF-treated samples were soaked in 250 ml of L1 water suspension (108 cells/ml) for different times at 25 C. A commercial deep-fat fryer was used for the frying tests. Fried potato crisp samples were analysed for moisture, asparagine content, colour, texture, oil uptake and acrylamide content. Results showed a significant reduction of asparagine content and therefore of acrylamide formation after frying. PEF pre-treatment of raw potato slices allowed the cell de- compartmentalisation enhancing the sugars and asparagine availability and so their assimilability from yeasts.

© 2018 The authors and IOS Press. Breast density is a crucial factor to follow-up the relapse of breast cancer in mammograms and the risk of local recurrence after conservative surgery and/or radiotherapy. Accurate breast density estimation with visual assessment is still a challenge due to faint contrast and significant variations in background fatty tissues in mammograms. The important key of breast density estimation is to properly detect the dense tissues in a mammographic image. Thus, this paper presents an automatic deep breast density segmentation using conditional Generative Adversarial Networks (cGAN) that consist of two successive deep networks: generator and discriminator. The generator network learns the mapping from the input mammogram to the output binary mask detection the area of the dense tissues. In turn, the discriminator learns a loss function to train this mapping by comparing the ground-truth and the predicted mask under observing the input mammogram as a condition. The performance of the proposed model was evaluated on the public INbreast mammographic datasets. The proposed model can segment the dense regions with overall recall, precision and F-score about 95%, 92%, and 93%, respectively, outperforming state-of-the-art of breast density segmentation. The proposed model can segment more than 40 images with a size of 512×512 per second on a recent GPU.

© Springer Nature Switzerland AG 2018. This paper proposes a novel approach based on conditional Generative Adversarial Networks (cGAN) for breast mass segmentation in mammography. We hypothesized that the cGAN structure is well-suited to accurately outline the mass area, especially when the training data is limited. The generative network learns intrinsic features of tumors while the adversarial network enforces segmentations to be similar to the ground truth. Experiments performed on dozens of malignant tumors extracted from the public DDSM dataset and from our in-house private dataset confirm our hypothesis with very high Dice coefficient and Jaccard index (>94% and >89%, respectively) outperforming the scores obtained by other state-of-the-art approaches. Furthermore, in order to detect portray significant morphological features of the segmented tumor, a specific Convolutional Neural Network (CNN) have also been designed for classifying the segmented tumor areas into four types (irregular, lobular, oval and round), which provides an overall accuracy about 72% with the DDSM dataset.

© 2018 The authors and IOS Press. This paper proposed a retinal image segmentation method based on conditional Generative Adversarial Network (cGAN) to segment optic disc. The proposed model consists of two successive networks: generator and discriminator. The generator learns to map information from the observing input (i.e., retinal fundus color image), to the output (i.e., binary mask). Then, the discriminator learns as a loss function to train this mapping by comparing the ground-truth and the predicted output with observing the input image as a condition. Experiments were performed on two publicly available dataset; DRISHTI GS1 and RIM-ONE. The proposed model outperformed state-of-the-art-methods by achieving around 0.96 and 0.98 of Jaccard and Dice coefficients, respectively. Moreover, an image segmentation is performed in less than a second on recent GPU.

© Springer Nature Switzerland AG 2018. Skin lesion segmentation (SLS) in dermoscopic images is a crucial task for automated diagnosis of melanoma. In this paper, we present a robust deep learning SLS model represented as an encoder-decoder network. The encoder network is constructed by dilated residual layers, in turn, a pyramid pooling network followed by three convolution layers is used for the decoder. Unlike the traditional methods employing a cross-entropy loss, we formulated a new loss function by combining both Negative Log Likelihood (NLL) and End Point Error (EPE) to accurately segment the boundaries of melanoma regions. The robustness of the proposed model was evaluated on two public databases: ISBI 2016 and 2017 for skin lesion analysis towards melanoma detection challenge. The proposed model outperforms the state-of-the-art methods in terms of the segmentation accuracy. Moreover, it is capable of segmenting about 100 images of a 384 × 384 size per second on a recent GPU.