This work presents sPyNNaker 4.0.0, the latest version of the software package for simulating PyNN-defined spiking neural networks (SNNs) on the SpiNNaker neuromorphic platform. Operations underpinning realtime SNN execution are presented, including an event-based operating system facilitating efficient time-driven neuron state updates and pipelined event-driven spike processing. Preprocessing, realtime execution, and neuron/synapse model implementations are discussed, all in the context of a simple example SNN. Simulation results are demonstrated, together with performance profiling providing insights into how software interacts with the underlying hardware to achieve realtime execution. System performance is shown to be within a factor of 2 of the original design target of 10,000 synaptic events per millisecond, however SNN topology is shown to influence performance considerably. A cost model is therefore developed characterizing the effect of network connectivity and SNN partitioning. This model enables users to estimate SNN simulation performance, allows the SpiNNaker team to make predictions on the impact of performance improvements, and helps demonstrate the continued potential of the SpiNNaker neuromorphic hardware.

BackgroundSynovitis occurring frequently in osteoarthritis (OA) may be a targeted outcome. There are no data examining whether synovitis changes following intra-articular intervention.MethodsPersons aged 40 years and older with painful knee OA participated in an open label trial of intra-articular steroid therapy. At all time points they completed the Knee Injury and Osteoarthritis Outcome Score (KOOS) questionnaire. They had a contrast-enhanced (CE) MRI immediately prior to an intra-articular steroid injection with a repeat scan within 20 days. Response status was assessed using the Osteoarthritis Research Society International (OARSI) response criteria. OARSI responders were followed until their pain relapsed either within 20% of baseline or 6 months, shortly after which a third MRI was performed. Synovial tissue volume (STV) was measured on postcontrast knee images. We looked at changes in the STV and in pain, and their association.Results120 subjects with preinjection and postinjection CE MRI were followed. Their mean age was 62.3 years (SD=10.3) and 62 (52%) were women. The median time between injection and follow-up scan was 8 days (IQR 7–14 days). 85/120 (71%) were OARSI responders. Pain decreased (mean change in KOOS=+23.9; 95% CI 20.1 to 27.8, p<0.001) following steroid injection, as did mean STV (mean change=−1071 mm3; 95% CI −1839 mm3 to −303 mm3, p=0.01). Of the 80 who returned for a third MRI, pain relapsed in 57, and in the 48 of those with MRI data, STV increased between follow-up and final visit (+1220 mm3; 95% CI 25 mm3 to 2414 mm3, p=0.05). 23 were persistent responders at 6 months and, in these, STV did not increase (mean change=−202 mm3; 95% CI −2008 mm3 to 1604 mm3, p=0.83). Controlling for variation over time, there was a significant association between synovitis volume and KOOS pain (b coefficient—change in KOOS pain score per 1000 mm3 change in STV=−1.13; 95% CI −1.87 to −0.39, p=0.003), although STV accounted for only a small proportion of the variance in change in pain.ConclusionsSynovial tissue volume in knee OA shrinks following steroid therapy, and rebounds in those whose pain relapses. It can be considered a treatment target in symptomatic knee OA.Trial registration numberISRCTN07329370.

ObjectiveBraces used to treat (PF) osteoarthritis (OA) may reduce contact stress across the PF joint. We hypothesised that in PF OA, braces would decrease knee pain and shrink PF bone marrow lesions (BMLs).MethodsEligible subjects had painful PF OA. Subjects were randomly allocated to brace or no brace for 6 weeks. Knee MRIs were acquired at baseline and 6 weeks. We measured BMLs on post-contrast fat suppressed sagittal and proton density weighted axial images. The primary symptom outcome was change in pain at 6 weeks during a preselected painful activity, and the primary structural outcome was BML volume change in the PF joint. Analyses used multiple linear regression.ResultsWe randomised 126 subjects aged 40–70 years (mean age 55.5  years; 72 females (57.1%)). Mean nominated visual analogue scale (0–10 cm) pain score at baseline was 6.5 cm. 94 knees (75%) had PF BMLs at baseline. Subjects wore the brace for a mean of 7.4 h/day. 6 subjects withdrew during the trial. After accounting for baseline values, the brace group had lower knee pain than the control group at 6 weeks (difference between groups −1.3 cm, 95% CI −2.0 to −0.7; p<0.001) and reduced PF BML volume (difference −490.6 mm3, 95% CI −929.5 to −51.7; p=0.03) but not tibiofemoral volume (difference −53.9 mm3, 95% CI −625.9 to 518.2; p=0.85).ConclusionsA PF brace reduces BML volume in the targeted compartment of the knee, and relieves knee pain.Trial registration numberUK. ISRCTN50380458.

Real-time simulation of a large-scale biologically representative spiking neural network is presented, through the use of a heterogeneous parallelization scheme and SpiNNaker neuromorphic hardware. A published cortical microcircuit model is used as a benchmark test case, representing ≈1 mm 2 of early sensory cortex, containing 77 k neurons and 0.3 billion synapses. This is the first hard real-time simulation of this model, with 10 s of biological simulation time executed in 10 s wall-clock time. This surpasses best-published efforts on HPC neural simulators (3 × slowdown) and GPUs running optimized spiking neural network (SNN) libraries (2 × slowdown). Furthermore, the presented approach indicates that real-time processing can be maintained with increasing SNN size, breaking the communication barrier incurred by traditional computing machinery. Model results are compared to an established HPC simulator baseline to verify simulation correctness, comparing well across a range of statistical measures. Energy to solution and energy per synaptic event are also reported, demonstrating that the relatively low-tech SpiNNaker processors achieve a 10 × reduction in energy relative to modern HPC systems, and comparable energy consumption to modern GPUs. Finally, system robustness is demonstrated through multiple 12 h simulations of the cortical microcircuit, each simulating 12 h of biological time, and demonstrating the potential of neuromorphic hardware as a neuroscience research tool for studying complex spiking neural networks over extended time periods. This article is part of the theme issue ‘Harmonizing energy-autonomous computing and intelligence’.

SummaryObjectivesKnee osteoarthritis (OA) is thought to be a slowly evolving disease with glacial changes in cartilage morphology necessitating trials of potential treatments lasting 1–2 years with evidence that over 6 months change in cartilage is not detectable. In contrast to cartilage, bone has the capacity to adapt rapidly, such as after fracture. We tested whether bone marrow lesions (BMLs) change in volume in 6 and 12 weeks, suggesting they may provide evidence of short term fluctuations of joint damage.MethodsIn 62 patients with patellofemoral knee OA (mean age 55.7 years, 59.7% women, mean BMI 31.0), we obtained baseline, 6 and 12 week knee MRIs with contrast enhancement. Of those with BMLs at baseline, we assessed BML volume on the axial proton density fat saturated (FS) images and postcontrast sagittal T1 weighted FS images. We manually segmented BML volumes, testing repeatability of BML volumes in knees remeasured. Using the standard deviation of the difference between repeated measurements to calculate Bland–Altman Limits of Agreement, we determined how much BML volume change represented a change greater than due to chance.ResultsFifty-two patients had BMLs at baseline. Test–retest reliability for BML volume was high (ICC 0.89, 95% CI 0.80–0.97). All knees showed at least some change in BML volume by 6 and 12 weeks. On the axial view at 6 weeks, 20/49 (40.8%) knees showed BML volume changes greater than the limits of agreement with similar results at 12 weeks. BML changes were evenly divided among knees with enlarging and shrinking BMLs. 63.3% of the knees had more than 50% change in BML volume at either 6 or 12 weeks on the axial view and 48.7% on the sagittal view.ConclusionsKnee BML volumes change in several weeks in many persons with knee OA. To the extent that they could be regarded as treatment targets, trials testing BML effects could avoid the usual prolonged structure modification trials.