We investigate the samples derived from generative adversarial networks (GAN) from a classification perspective. We train a classifier on generated samples and on real data and see how they compared on a held out validation set. We see that recent GAN models which produce visually convincing samples are not yet able to match the training on real data. To analyse this we compare training a classifier on generated samples and various sizes of the real training set. We propose architectural and algorithmic changes to reduce this gap. First, we show that a modification to the GAN architecture is needed, which leads to improve generation of samples. Second, we use multiple GAN models as a way to cover the real data distribution, again leading to improvement in classifier training. We also show that in the case of training on small number of samples, a GAN model provides better compression in terms of storage requirements as compared to the real data.
Objectives: The study aimed to evaluate clinical and paraclinical characteristics of left atrial myxoma and the early result of minimally invasive left atrial myxoma resection at E hospital.
Patients and methods: This is retrospective, descriptive study of consecutive patients, who underwent minimally invasive left atrial myxoma rejection, using total or video-assisted endoscopic technique from October 2016 to March 2021 at E hospital. There were 31 patients, consisting of 27 females and 04 males. The mean age was 53 ± 13 years old (range [17-74]).
Results: Preoperative clinical manifestations were diverse. Asymptomatic form was in 3 patients (9.7%), hemodynamic symptoms were in 26 patients (83.9%), embolism were presented in 4 patients (12.9%), systemic symptoms were in 8 patients (25.8%). Anemia and elevated erythrocyte sedimentation rates were observed in 45.2% and 74.2%, respectively. Echocardiography results: the average tumor size was 4.2 ± 1.7 cm (range [1.7-8]), the site of attachment was mainly in the atrial septum (77.4%). Cardiopulmonary bypass time was 158 ± 43 minutes (range [100-252]), cross-clamp time was 84 ± 34.2 minutes (range [42-153]), ventilation time was 10,8 ± 7,0 hours (range [3-30]), intensive care unit stay was 1,5 ± 1,0 days (range [0.5-4]), and in-hospital stay was 9.5 ± 5.0 days (range [3-30]). There was no hospital mortality. Cerebrovascular accident was presented in one (3.2%), femoral artery stenosis was in one (3.2%), atrial fibrillation after surgery was in one(3.2%).There was no bleeding, that require reoperation, and no other serious complications.
Conclusions: Initial results of left atrial myxoma resection, using minimally invasive total or video-assisted technique wassafe and effectivewith low complications, could be recommended toapply routinely in cardiac surgery centers.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.