Storage and Learning Phase Transitions in the Random-Features Hopfield Model

The Hopfield model is a paradigmatic model of neural networks that has been analyzed for many decades in the statistical physics, neuroscience, and machine learning communities. Inspired by the manifold hypothesis in machine learning, we propose and investigate a generalization of the standard setting that we name random-features Hopfield model. Here, P binary patterns of length N are generated by applying to Gaussian vectors sampled in a latent space of dimension D a random projection followed by a nonlinearity. Using the replica method from statistical physics, we derive the phase diagram of the model in the limit P,N,D→∞ with fixed ratios α=P/N and α_{D}=D/N. Besides the usual retrieval phase, where the patterns can be dynamically recovered from some initial corruption, we uncover a new phase where the features characterizing the projection can be recovered instead. We call this phenomena the learning phase transition, as the features are not explicitly given to the model but rather are inferred from the patterns in an unsupervised fashion.

General Anesthesia Versus Conscious Sedation and Local Anesthesia During Thrombectomy for Acute Ischemic Stroke

Background and Purpose:

As numerous questions remain about the best anesthetic strategy during thrombectomy, we assessed functional and radiological outcomes in stroke patients treated with thrombectomy in presence of general anesthesia (GA) versus conscious sedation (CS) and local anesthesia (LA).

Methods:

We conducted a cohort study on prospectively collected data from 4429 patients enrolled in the Italian Registry of Endovascular Treatment in Acute Stroke.

Results:

GA was used in 2013 patients, CS in 1285 patients, and LA in 1131 patients. The rates of 3-month modified Rankin Scale score of 0–1 were 32.7%, 33.7%, and 38.1% in the GA, CS, and LA groups: GA versus CS: odds ratios after adjustment for unbalanced variables (adjusted odds ratio [aOR]), 0.811 (95% CI, 0.602–1.091); and GA versus LA: aOR, 0.714 (95% CI, 0.515–0.990). The rates of modified Rankin Scale score of 0–2 were 42.5%, 46.6%, and 52.4% in the GA, CS, and LA groups: GA versus CS: aOR, 0.902 (95% CI, 0.689–1.180); and GA versus LA: aOR, 0.769 (95% CI, 0.566–0.998). The rates of 3-month death were 21.5%, 19.7%, and 14.8% in the GA, CS, and LA groups: GA versus CS: aOR, 0.872 (95% CI, 0.644–1.181); and GA versus LA: aOR, 1.235 (95% CI, 0.844–1.807). The rates of parenchymal hematoma were 9%, 12.6%, and 11.3% in the GA, CS, and LA groups: GA versus CS: aOR, 0.380 (95% CI, 0.262–0.551); and GA versus LA: aOR, 0.532 (95% CI, 0.337–0.838). After model of adjustment for predefined variables (age, sex, thrombolysis, National Institutes of Health Stroke Scale, onset-to-groin time, anterior large vessel occlusion, procedure time, prestroke modified Rankin Scale score of <1, antiplatelet, and anticoagulant), differences were found also between GA versus CS as regards modified Rankin Scale score of 0–2 (aOR, 0.659 [95% CI, 0.538–0.807]) and GA versus LA as regards death (aOR, 1.413 [95% CI, 1.095–1.823]).

Conclusions:

GA during thrombectomy was associated with worse 3-month functional outcomes, especially when compared with LA. The inclusion of an LA arm in future randomized clinical trials of anesthesia strategy is recommended.

Flow diverters for ruptured posterior circulation perforator aneurysms: multicenter experience and literature review

BACKGROUND

The management of ruptured posterior circulation perforator aneurysms (rPCPAs) remains unclear. We present our experience in treating rPCPAs with flow diverter stents (FDs) and evaluate their safety and efficacy at mid- to long-term follow-up. A diagnostic and therapeutic algorithm for rPCPAs is also proposed.

METHODS

We retrospectively analyzed data from all consecutive patients with rPCPAs treated with FDs at our institutions between January 2013 and July 2019. Clinical presentations, time of treatments, intra- and perioperative complications, and clinical and angiographic outcomes were recorded, with a mid- to long-term follow-up. A systematic review of the literature on rPCPAs treated with FDs was also performed.

RESULTS

Seven patients with seven rPCPAs were treated with FDs. All patients presented with an atypical subarachnoid hemorrhage distribution and a low to medium Hunt-Hess grade. In 29% of cases rPCPAs were identified on the initial angiogram. In 57% of cases, FDs were inserted within 2 days of the diagnosis. Immediate aneurysm occlusion was observed in 14% of the cases and in 71% at the first follow-up (mean 2.4 months). At mean follow-up of 33 months (range 3-72 months) one case of delayed ischemic complication occurred. Six patients had a modified Rankin Scale (mRS) score of 0 and one patient had an mRS score of 4 at the latest follow-up.

CONCLUSIONS

The best management for rPCPAs remains unclear, but FDs seem to have lower complication rates than other treatment options. Further studies with larger series are needed to confirm the role of FDs in rPCPA.

Long-term follow-up in the endovascular treatment of intracranial aneurysms with flow-diverter stents: update of a single-centre experience

INTRODUCTION

Flow-diverter stents are becoming a useful tool in treating patients with intracranial aneurysms with suitable anatomical feature. Purpose of this study was to evaluate effectiveness and safety of endovascular treatment with flow-diverting stents (FD) in unruptured intracranial aneurysms.

METHODS

From May 2009 and May 2014, we treated 49 patients with a total of 58 aneurysms, with FD technique. All patients were treated electively, under general anesthesia and were administered single antiplatelet drug 5 days before the procedure and double antiplatelet therapy for 3 months afterwards. Fifteen of the patients were asymptomatic, eight had headache, thirteen patients presented symptoms due to mass effect of the aneurysm on CNS structures, twelve were treated due to a post-surgical relapse and one patient presented relapsing TIAs due to distal embolization from the aneurysm dome. Choice of FD treatment was done according to aneurysm anatomy (fusiform over saccular, dome/neck ratio < 2) and whenever conventional treatment (coil embolization) appeared difficult (eg. Large aneurysm neck, fusiform aneurysms or difficult sac catheterization). We considered a dome/neck ratio > 2 as the only exclusion criteria.

RESULTS

Successful stent deployment was achieved in 50 procedures out of 52 (94.34%) while overall mortality was 2% (1/49). Forty-eight patients were evaluated at long-term follow-up for a total of 56 treated aneurysms. At 3 months, follow-up 75% (42/56) of the aneurysms were excluded from intracranial circulation, at 6 months 80.35% (45/56) and at 12 months 84% (47/56). Stent patency was observed in 100% of patients at short and long-term follow-up, with only two cases of intimal hyperplasia at 3 months, without any further complications.

CONCLUSIONS

According to our study FD repair of unruptured intracranial aneurysms appeared to be a safe and effective technique, especially in selected patients with hostile anatomy for traditional embolization.

Endovascular treatment of intracranial aneurysms with flow-diverter stents: preliminary single-centre experience

PURPOSE

This paper reports our preliminary experience with the endovascular treatment of intracranial aneurysms using flow-diverter stents (FDs) and compares it with the literature data.

MATERIALS AND METHODS

From May 2009 to April 2012, 28 patients (6 men and 22 women; mean age, 54 years) with a total of 35 aneurysms were treated with FDs. We evaluated postprocedural technical success and long-term efficacy, with follow-up examinations performed at 3-7 days [computed tomography (CT)/magnetic resonance (MR) angiography] and at 3, 6 and 12 months (digital subtraction angiography, DSA). A total of 43 FDs were placed, 36 Pipeline and 7 Silk.

RESULTS

A total of 30 procedures were performed (two patients were treated twice). Technical success was 96.6%, with one case of postprocedural death; the aneurysm exclusion rate at 3, 6 and 12 months was 60%, 73% and 89%, respectively. There was no case of acute stent thrombosis, and only two cases of nonsignificant stenosis. All covered side branches were patent, except one case of steno-occlusion of the ophthalmic artery.

CONCLUSIONS

Our results are consistent with the literature and demonstrate the effectiveness and safety of FDs in selected cases of cerebral aneurysm (wide neck, fusiform, blister-like).