Critical analysis on the reproducibility of visual quality assessment using deep features.

Data used to train supervised machine learning models are commonly split into independent training, validation, and test sets. This paper illustrates that complex data leakage cases have occurred in the no-reference image and video quality assessment literature. Recently, papers in several journals reported performance results well above the best in the field. However, our analysis shows that information from the test set was inappropriately used in the training process in different ways and that the claimed performance results cannot be achieved. When correcting for the data leakage, the performances of the approaches drop even below the state-of-the-art by a large margin. Additionally, we investigate end-to-end variations to the discussed approaches, which do not improve upon the original.

Blind Quality Assessment of Iris Images Acquired in Visible Light for Biometric Recognition.

Image quality is a key issue affecting the performance of biometric systems. Ensuring the quality of iris images acquired in unconstrained imaging conditions in visible light poses many challenges to iris recognition systems. Poor-quality iris images increase the false rejection rate and decrease the performance of the systems by quality filtering. Methods that can accurately predict iris image quality can improve the efficiency of quality-control protocols in iris recognition systems. We propose a fast blind/no-reference metric for predicting iris image quality. The proposed metric is based on statistical features of the sign and the magnitude of local image intensities. The experiments, conducted with a reference iris recognition system and three datasets of iris images acquired in visible light, showed that the quality of iris images strongly affects the recognition performance and is highly correlated with the iris matching scores. Rejecting poor-quality iris images improved the performance of the iris recognition system. In addition, we analyzed the effect of iris image quality on the accuracy of the iris segmentation module in the iris recognition system.

KonIQ-10k: An ecologically valid database for deep learning of blind image quality assessment.

Deep learning methods for image quality assessment (IQA) are limited due to the small size of existing datasets. Extensive datasets require substantial resources both for generating publishable content and annotating it accurately. We present a systematic and scalable approach to creating KonIQ-10k, the largest IQA dataset to date, consisting of 10,073 quality scored images. It is the first in-the-wild database aiming for ecological validity, concerning the authenticity of distortions, the diversity of content, and quality-related indicators. Through the use of crowdsourcing, we obtained 1.2 million reliable quality ratings from 1,459 crowd workers, paving the way for more general IQA models. We propose a novel, deep learning model (KonCept512), to show an excellent generalization beyond the test set (0.921 SROCC), to the current state-of-the-art database LIVE-in-the-Wild (0.825 SROCC). The model derives its core performance from the InceptionResNet architecture, being trained at a higher resolution than previous models (512 × 384). Correlation analysis shows that KonCept512 performs similar to having 9 subjective scores for each test image.

Kinetic analysis of oxygen dynamics under a variable work rate.

Measurements of oxygen uptake are central to methods for the assessment of physical fitness and endurance capabilities in athletes. Two important parameters extracted from such data of incremental exercise tests are the maximal oxygen uptake and the critical power. A commonly accepted model of the dynamics of oxygen uptake during exercise at a constant work rate comprises a constant baseline oxygen uptake, an exponential fast component, and another exponential slow component for heavy and severe work rates. We have generalized this model to variable load protocols with differential equations that naturally correspond to the standard model for a constant work rate. This provides the means for predicting the oxygen uptake response to variable load profiles including phases of recovery. The model parameters have been fitted for individual subjects from a cycle ergometer test, including the maximal oxygen uptake and critical power. The model predictions have been validated by data collected in separate tests. Our findings indicate that the oxygen kinetics for a variable exercise load can be predicted using the generalized mathematical standard model. Such models can be applied in the field where the constant work rate assumption generally is not valid.

Bicycling suppresses abnormal beta synchrony in the Parkinsonian basal ganglia.

OBJECTIVE: Freezing of gait is a poorly understood symptom of Parkinson disease, and can severely disrupt the locomotion of affected patients. However, bicycling ability remains surprisingly unaffected in most patients suffering from freezing, suggesting functional differences in the motor network. The purpose of this study was to characterize and contrast the oscillatory dynamics underlying bicycling and walking in the basal ganglia. METHODS: We present the first local field potential recordings directly comparing bicycling and walking in Parkinson disease patients with electrodes implanted in the subthalamic nuclei for deep brain stimulation. Low (13-22Hz) and high (23-35Hz) beta power changes were analyzed in 22 subthalamic nuclei from 13 Parkinson disease patients (57.5 ± 5.9 years old, 4 female). The study group consisted of 5 patients with and 8 patients without freezing of gait. RESULTS: In patients without freezing of gait, both bicycling and walking led to a suppression of subthalamic beta power (13-35Hz), and this suppression was stronger for bicycling. Freezers showed a similar pattern in general. Superimposed on this pattern, however, we observed a movement-induced, narrowband power increase around 18Hz, which was evident even in the absence of freezing. INTERPRETATION: These results indicate that bicycling facilitates overall suppression of beta power. Furthermore, movement leads to exaggerated synchronization in the low beta band specifically within the basal ganglia of patients susceptible to freezing. Abnormal ∼18Hz oscillations are implicated in the pathophysiology of freezing of gait, and suppressing them may form a key strategy in developing potential therapies. Ann Neurol 2017;82:592-601.

Bicycling and Walking are Associated with Different Cortical Oscillatory Dynamics.

Although bicycling and walking involve similar complex coordinated movements, surprisingly Parkinson's patients with freezing of gait typically remain able to bicycle despite severe difficulties in walking. This observation suggests functional differences in the motor networks subserving bicycling and walking. However, a direct comparison of brain activity related to bicycling and walking has never been performed, neither in healthy participants nor in patients. Such a comparison could potentially help elucidating the cortical involvement in motor control and the mechanisms through which bicycling ability may be preserved in patients with freezing of gait. The aim of this study was to contrast the cortical oscillatory dynamics involved in bicycling and walking in healthy participants. To this end, EEG and EMG data of 14 healthy participants were analyzed, who cycled on a stationary bicycle at a slow cadence of 40 revolutions per minute (rpm) and walked at 40 strides per minute (spm), respectively. Relative to walking, bicycling was associated with a stronger power decrease in the high beta band (23-35 Hz) during movement initiation and execution, followed by a stronger beta power increase after movement termination. Walking, on the other hand, was characterized by a stronger and persisting alpha power (8-12 Hz) decrease. Both bicycling and walking exhibited movement cycle-dependent power modulation in the 24-40 Hz range that was correlated with EMG activity. This modulation was significantly stronger in walking. The present findings reveal differential cortical oscillatory dynamics in motor control for two types of complex coordinated motor behavior, i.e., bicycling and walking. Bicycling was associated with a stronger sustained cortical activation as indicated by the stronger high beta power decrease during movement execution and less cortical motor control within the movement cycle. We speculate this to be due to the more continuous nature of bicycling demanding less phase-dependent sensory processing and motor planning, as opposed to walking.

BrainCycles: Experimental Setup for the Combined Measurement of Cortical and Subcortical Activity in Parkinson's Disease Patients during Cycling.

Recently, it has been demonstrated that bicycling ability remains surprisingly preserved in Parkinson's disease (PD) patients who suffer from freezing of gait. Cycling has been also proposed as a therapeutic means of treating PD symptoms, with some preliminary success. The neural mechanisms behind these phenomena are however not yet understood. One of the reasons is that the investigations of neuronal activity during pedaling have been up to now limited to PET and fMRI studies, which restrict the temporal resolution of analysis, and to scalp EEG focused on cortical activation. However, deeper brain structures like the basal ganglia are also associated with control of voluntary motor movements like cycling and are affected by PD. Deep brain stimulation (DBS) electrodes implanted for therapy in PD patients provide rare and unique access to directly record basal ganglia activity with a very high temporal resolution. In this paper we present an experimental setup allowing combined investigation of basal ganglia local field potentials (LFPs) and scalp EEG underlying bicycling in PD patients. The main part of the setup is a bike simulator consisting of a classic Dutch-style bicycle frame mounted on a commercially available ergometer. The pedal resistance is controllable in real-time by custom software and the pedal position is continuously tracked by custom Arduino-based electronics using optical and magnetic sensors. A portable bioamplifier records the pedal position signal, the angle of the knee, and the foot pressure together with EEG, EMG, and basal ganglia LFPs. A handlebar-mounted display provides additional information for patients riding the bike simulator, including the current and target pedaling rate. In order to demonstrate the utility of the setup, example data from pilot recordings are shown. The presented experimental setup provides means to directly record basal ganglia activity not only during cycling but also during other movement tasks in patients who have undergone DBS treatment. Thus, it can facilitate studies comparing bicycling and walking, to elucidate why PD patients often retain the ability to bicycle despite severe freezing of gait. Moreover it can help clarifying the mechanism through which cycling may have therapeutic benefits.