Mode combinability: Exploring convex combinations of permutation aligned models.

We explore element-wise convex combinations of two permutation-aligned neural network parameter vectors ΘA and ΘB of size d. We conduct extensive experiments by examining various distributions of such model combinations parametrized by elements of the hypercube [0,1]d and its vicinity. Our findings reveal that broad regions of the hypercube form surfaces of low loss values, indicating that the notion of linear mode connectivity extends to a more general phenomenon which we call mode combinability. We also make several novel observations regarding linear mode connectivity and model re-basin. We demonstrate a transitivity property: two models re-based to a common third model are also linear mode connected, and a robustness property: even with significant perturbations of the neuron matchings the resulting combinations continue to form a working model. Moreover, we analyze the functional and weight similarity of model combinations and show that such combinations are non-vacuous in the sense that there are significant functional differences between the resulting models.

Novel electrocardiographic dyssynchrony criteria improve patient selection for cardiac resynchronization therapy.

Aims: We hypothesized that the greater the intra- or interventricular dyssynchrony (intraD, interD), the more effective cardiac resynchronization therapy (CRT) is. We sought to improve patient selection for CRT by using novel ECG dyssynchrony criteria. Methods and results: Left ventricular (LV) intraD was estimated by the absolute time difference between the intrinsicoid deflections (ID) in leads aVL and aVF divided by the QRS duration (QRSd): [aVLID - aVFID]/QRSd (%). InterD was estimated from the formula: [V5ID - V1ID]/QRSd (%). Their >25% value indicated electrical dyssynchrony present (ED+) and ≤25% value electrical dyssynchrony absent (ED-) diagnoses. Using the intraD + interD criteria (intra + interDC) together, if at least one of them indicated ED+ diagnosis, a final ED+ diagnosis, if both indicated ED- diagnosis, a final ED- diagnosis was made. Two authors, blinded to CRT response, retrospectively analysed pre-CRT ECGs of 124 patients with known CRT outcome. CRT response was defined as improvement of ≥ 1 NYHA class, being alive and having no hospitalizations for heart failure during 6 months of follow-up. 35/124 (28%) patients were non-responders (NRs), using the traditional criteria (TC) correct diagnosis was made in the remaining 89/124 (72%) responder (R) cases. The test accuracy (TA) of intra + interDC + TC [100/124 (81%), P < 0.001] was superior to that of TC [89/124 (72%)] due to its superior TA [36/43 (84%) vs. 29/43 (67%), respectively, P = 0.0156] in the non-specific intra-ventricular conduction disturbance (NICD) subgroup [43/124 (35%)]. In the left bundle branch block subgroup [70/124 (56%)] there was no between-criteria difference in TA. Conclusion: The intra + interDC + TC predicts clinical response after CRT more accurately than TC alone, due to greater TA in the NICD subgroup.

The mechanism of reduced longitudinal left ventricular systolic function in hypertensive patients with normal ejection fraction.

BACKGROUND: MacIver and Townsend's hypothesis predicts, based on a mathematical model of left ventricular contraction, that preserved absolute radial wall thickening (radWT) due to left ventricular hypertrophy is responsible for the normal ejection fraction in patients with heart failure with preserved ejection fraction (HFPEF). METHODS: We tested the validity of this hypothesis by detailed echocardiography including evaluation of ventricular myocardial strain (S) using speckle tracking imaging in at least 60-year-old 18 controls and 94 hypertensive patients with normal ejection fraction. RESULTS: Echocardiography revealed no left ventricular diastolic dysfunction in 38 out of 94 (40%) patients with hypertension (HTDD-negative group), and 56 out of 94 (60%) patients had diastolic dysfunction (HTDD-positive groups). The absolute values of global longitudinal left ventricular peak systolic S were significantly reduced in both patient groups (P < 0.05 for HTDD-negative, P < 0.01 for HTDD-positive groups) vs. the controls. There were no significant between-groups differences in circumferential and radial peak left ventricular systolic Ss, radWT and ejection fraction. Left ventricular mass (LVM) (P < 0.001), LVM/BMI (P < 0.01) increased in the HTDD-positive group and ejection fraction/LVM/BMI decreased in both patient groups (P < 0.01 for HTDD-negative, P < 0.001 for HTDD-positive groups) vs. the controls. LVM increased, ejection fraction/LVM/BMI decreased in the HTDD-positive group vs. the HTDD-negative group (P < 0.05 and P < 0.01, respectively). CONCLUSION: We demonstrated decreased longitudinal left ventricular systolic function and showed that preserved ejection fraction was due to preserved absolute radWT and not due to increased radial or circumferential systolic function in patients with hypertension and normal ejection fraction, a potential HFPEF precursor condition. Instead of ejection fraction, rather ejection fraction/LVM/BMI might be used to detect subtle left ventricular systolic dysfunction in hypertension and HFPEF.

Inflammation and oxidative stress caused by nitric oxide synthase uncoupling might lead to left ventricular diastolic and systolic dysfunction in patients with hypertension.

OBJECTIVE: To investigate the role of oxidative stress, inflammation, hypercoagulability and neuroendocrine activation in the transition of hypertensive heart disease to heart failure with preserved ejection fraction (HFPEF). METHODS: We performed echocardiography for 112 patients (≥ 60 years old) with normal EF (18 controls and 94 with hypertension), and determined protein carbonylation (PC), and tetrahydrobiopterin (BH4), C-reactive protein (CRP), interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), fibrinogen, plasminogen activator inhibitor type-I (PAI-I), von Willebrand factor, chromogranin A (cGA) and B-type natriuretic peptide (BNP) levels from their blood samples. RESULTS: We found that 40% (38/94) of the patients with hypertension (HT) had no diastolic dysfunction (HTDD-), and 60% (56/94) had diastolic dysfunction (HTDD+). Compared to the controls, both patient groups had increased PC and BH4, TNF-α, PAI-I and BNP levels, while the HTDD+ group had elevated cGA and CRP levels. Decreased atrial and longitudinal left ventricular (LV) systolic and diastolic myocardial deformation (strain and strain rate) was demonstrated in both patient groups versus the control. Patients whose LV diastolic function deteriorated during the follow-up had elevated PC and IL-6 level compared to their own baseline values, and to the respective values of patients whose LV diastolic function remained unchanged. Oxidative stress, inflammation, BNP and PAI-I levels inversely correlated with LV systolic, diastolic and atrial function. CONCLUSIONS: In patients with HT and normal EF, the most common HFPEF precursor condition, oxidative stress and inflammation may be responsible for LV systolic, diastolic and atrial dysfunction, which are important determinants of the transition of HT to HFPEF.

Poly(I:C)-mediated tumor growth suppression in EGF-receptor overexpressing tumors using EGF-polyethylene glycol-linear polyethylenimine as carrier.

PURPOSE: To develop a novel polyethylenimine (PEI)-based polymeric carrier for tumor-targeted delivery of cytotoxic double-stranded RNA polyinosinic:polycytidylic acid, poly(I:C). The novel carrier should be chemically less complex but at least as effective as a previously developed tetra-conjugate containing epidermal growth factor (EGF) as targeting ligand, polyethylene glycol (PEG) as shielding spacer, 25 kDa branched PEI as RNA binding and endosomal buffering agent, and melittin as endosomal escape agent. METHODS: Novel conjugates were designed employing a simplified synthetic strategy based on 22 kDa linear polyethylenimine (LPEI), PEG spacers, and recombinant EGF. The efficacy of various conjugates (different PEG spacers, with and without targeting EGF) in poly(I:C)-mediated cell killing was evaluated in vitro using two human U87MG glioma cell lines. The most effective polyplex was tested for in vivo activity in A431 tumor xenografts. RESULTS: Targeting conjugate LPEI-PEG2 kDa-EGF was found as most effective in poly(I:C)-triggered killing of tumor cells in vitro. The efficacy correlated with glioma cell EGFR levels. Repeated intravenous administration of poly(I:C) polypexes strongly retarded growth of A431 human tumor xenograft in mice. CONCLUSIONS: The optimized LPEI-PEG2 kDa-EGF conjugate displays reduced chemical complexity and efficient poly(I:C)-mediated killing of EGFR overexpressing tumors in vitro and in vivo.