The lack of interpretability and transparency in deep learning architectures has raised concerns among professionals in various industries and academia. One of the main concerns is the ability to trust these architectures’ without being provided any insight into the decision-making process. Despite these concerns, researchers continue to explore new models and architectures that do not incorporate explainability into their main construct. In the medical industry, it is crucial to provide explanations of any decision, as patient health outcomes can vary according to decisions made. Furthermore, in medical research, incorrectly diagnosed neurological conditions are a high-cost error that contributes significantly to morbidity and mortality. Therefore, the development of new transparent techniques for neurological conditions is critical. This paper presents a novel Autonomous Relevance Technique for an Explainable neurological disease prediction framework called ART-Explain. The proposed technique autonomously extracts features from within the deep learning architecture to create novel visual explanations of the resulting prediction. ART-Explain is an end-to-end autonomous explainable technique designed to present an intuitive and holistic overview of a prediction made by a deep learning classifier. To evaluate the effectiveness of our approach, we benchmark it with other state-of-the-art techniques using three data sets of neurological disorders. The results demonstrate the generalisation capabilities of our technique and its suitability for real-world applications. By providing transparent insights into the decision-making process, ART-Explain can improve end-user trust and enable a better understanding of classification outcomes in the detection of neurological diseases.

Despite evidence indicating that black box models are problematic and not fully utilised across various domains, researchers persist in developing and employing these techniques without offering insights into their decision-making processes. As a result, in critical areas such as the medical domain, a lack of trust and scepticism persists. This paper introduces the OH-ART-Explain (Optimized Hybrid Model Selection for Autonomous Relevance Technique Explainable) framework as an enhanced version of the ART-Explain framework. The OH-ART-Explain framework utilises an optimisation algorithm called “Optimized Hybrid Model Selection” to identify the best deep learning model, feature extraction layer, and rule-based classification combination, leading to more robust results. Unlike the original ART-Explain framework, OH-ART-Explain does not limit itself to a single potential deep learning classifier, layer, or rule-based classification combination, thus improving flexibility and performance. Additionally, we propose the Reduced Graphical Explanation Technique (R-GET) to simplify the visualisation and interpretation of the rule-based classification process within the OH-ART-Explain framework. We evaluated the proposed framework using eight deep learning classifiers, two rule-based classifiers, and three medical datasets. Our experimentation result's key findings demonstrate that the synergy between the base model and rule-based classifier is interdependent. The combination within the OH-ART-Explain framework produces higher results than using only the deep learning classifier.

A group of highly experienced pilots performed full-motion, simulated T-38 landings before and after extended missions aboard the International Space Station (ISS). On the day of return from the ISS pilots’ performance was degraded on the initial landing attempt, with difficulty maintaining altitude during banking turns and navigational errors, which affected touchdown parameters such as touchdown speed, height over runway threshold and touchdown distance from the runway threshold. A positive result was that all pilots successfully completed a second landing attempt on the same day, suggesting a rapid recovery of performance once exposed to the task at hand. These results are consistent with a previous study that demonstrated significant deficits in post-flight driving performance, and both the pilot and driver subjects’ performance recovered to pre-flight levels within four days of return from the ISS. We propose that the primary factors underlying the post-flight performance deficits were the inability to respond appropriately to gravitational and visual tilts and a reduction in multitasking ability.

Deep neural networks (DNN) are a popular tool to process environmental sounds and identify sound-producing animals, but it can be difficult to understand the decision-making logic, particularly when it does not produce the expected results. Here we describe a new and enhanced visual interactive analysis of embeddings and explore its application in bioacoustics. Embeddings are the output of the penultimate layer of a DNN, an N-dimensional vector that, only one step removed from the final output, represent the inner-workings of a DNN model. Using existing dimensionality reduction techniques we converted the N-dimensional embeddings into 2 or 3-dimensional arrays displayed in scatterplots. By incorporating sound samples into the scatterplots we developed a visual and aural interactive interface and demonstrate its utility in assessing the performance of trained bioacoustic models, facilitating post-processing of results, error detection, input selection and the detection of rare events, which the reader can experience in online examples with publicly available code.

The otoliths of the vestibular system are seen as the primary gravitational sensors and are responsible for a compensatory eye torsion called the ocular counter-roll (OCR). The OCR ensures gaze stabilization and is sensitive to a lateral head roll with respect to gravity and the Gravito-Inertial Acceleration (GIA) vector during e.g., centrifugation. This otolith-mediated reflex will make sure you will still be able to maintain gaze stabilization and postural stability when making sharp turns during locomotion. To measure the effect of prolonged spaceflight on the otoliths, we measured the OCR induced by off-axis centrifugation in a group of 27 cosmonauts before and after their 6-month space mission to the International Space Station (ISS). We observed a significant decrease in OCR early post-flight, with first- time flyers being more strongly affected compared to frequent or experienced flyers. Our results strongly suggest that experienced space crew have acquired the ability to adapt faster after G-transitions and should therefore be sent for more challenging space missions, e.g., Moon or Mars, because they are noticeably less affected by microgravity regarding their vestibular system.

Otoliths are the primary gravity sensors of the vestibular system and are responsible for the ocular counter-roll (OCR). This compensatory eye torsion ensures gaze stabilization and is sensitive to a head roll with respect to gravity and the Gravito-Inertial Acceleration vector during, e.g., centrifugation. To measure the effect of prolonged spaceflight on the otoliths, we quantified the OCR induced by off-axis centrifugation in a group of 27 cosmonauts in an upright position before and after their 6-month space mission to the International Space Station. We observed a significant decrease in OCR early postflight, larger for first-time compared to experienced flyers. We also found a significantly larger torsion for the inner eye, the eye closest to the rotation axis. Our results suggest that experienced cosmonauts have acquired the ability to adapt faster after G-transitions. These data provide a scientific basis for sending experienced cosmonauts on challenging missions that include multiple g-level transitions.

The use of autonomous recordings of animal sounds to detect species is a popular conservation tool, constantly improving in fidelity as audio hardware and software evolves. Current classification algorithms utilise sound features extracted from the recording rather than the sound itself, with varying degrees of success. Neural networks that learn directly from the raw sound waveforms have been implemented in human speech recognition but the requirements of detailed labelled data have limited their use in bioacoustics. Here we test SincNet, an efficient neural network architecture that learns from the raw waveform using sinc-based filters. Results using an off-the-shelf implementation of SincNet on a publicly available bird sound dataset (NIPS4Bplus) show that the neural network rapidly converged reaching accuracies of over 65% with limited data. Their performance is comparable with traditional methods after hyperparameter tuning but they are more efficient. Learning directly from the raw waveform allows the algorithm to select automatically those elements of the sound that are best suited for the task, bypassing the onerous task of selecting feature extraction techniques and reducing possible biases. We use publicly released code and datasets to encourage others to replicate our results and to apply SincNet to their own datasets; and we review possible enhancements in the hope that algorithms that learn from the raw waveform will become useful bioacoustic tools.

Smart farming has become imperative these days due to competition, and use of Unmanned Aerial Vehicle (UAV) imagery is becoming an integral part of the process. Machine learning techniques have been successfully applied to capture UAV imagery of various spectral bands to identify weed infestations. Identification of weeds in chilli crop is a challenging task. In this paper, RGB images captured by drones have been used to detect weed in chilli field. This task has been addressed through orthomasaicking of images, feature extraction, labelling of images to train machine learning algorithms, and use of unsupervised learning with random forest for classification. MATLAB has been used for all computations and out-of-bag accuracy achieved for identifying weeds is 96 % .

To reach the goal of sustainable agriculture, smart farming is taking advantage of the Unmanned Aerial Vehicles (UAVs) and Internet of Things (IoT) paradigm. These smart farms are designed to be run by interconnected devices and vehicles. Some enormous potentials can be achieved by the integration of different IoT technologies to achieve automated operations with minimum supervision. This paper outlines some major applications of IoT and UAV in smart farming, explores the communication technologies, network functionalities and connectivity requirements for Smart farming. The connectivity limitations of smart agriculture and it's solutions are analysed with two case studies. In case study-1, we propose and evaluate meshed Long Range Wide Area Network (LoRaWAN) gateways to address connectivity limitations of Smart Farming. While in case study-2, we explore satellite communication systems to provide connectivity to smart farms in remote areas of Australia. Finally, we conclude the paper by identifying future research challenges on this topic and outlining directions to address those challenges.

This paper explores the potential of machine learning algorithms for weed and crop classification from UAV images. The identification of weeds in crops is a challenging task that has been addressed through orthomosaicing of images, feature extraction and labelling of images to train machine learning algorithms. In this paper, the performances of several machine learning algorithms, random forest (RF), support vector machine (SVM) and k-nearest neighbours (KNN), are analysed to detect weeds using UAV images collected from a chilli crop field located in Australia. The evaluation metrics used in the comparison of performance were accuracy, precision, recall, false positive rate and kappa coefficient. MATLAB is used for simulating the machine learning algorithms; and the achieved weed detection accuracies are 96% using RF, 94% using SVM and 63% using KNN. Based on this study, RF and SVM algorithms are efficient and practical to use, and can be implemented easily for detecting weed from UAV images.

Parkinson's disease (PD) is a neurodegenerative disorder associated with motor and non-motor symptoms. Current treatments primarily focus on managing motor symptom severity such as tremor, bradykinesia, and rigidity. However, as the disease progresses, treatment side-effects can emerge such as on/off periods and dyskinesia. The objective of the Levodopa Response Study was to identify whether wearable sensor data can be used to objectively quantify symptom severity in individuals with PD exhibiting motor fluctuations. Thirty-one subjects with PD were recruited from 2 sites to participate in a 4-day study. Data was collected using 2 wrist-worn accelerometers and a waist-worn smartphone. During Days 1 and 4, a portion of the data was collected in the laboratory while subjects performed a battery of motor tasks as clinicians rated symptom severity. The remaining of the recordings were performed in the home and community settings. To our knowledge, this is the first dataset collected using wearable accelerometers with specific focus on individuals with PD experiencing motor fluctuations that is made available via an open data repository.

The measurement of 3D eye position is an important investigative tool in the understanding of the human vestibular and oculomotor systems. The subject is asked to look at a fixation point positioned directly ahead. This eye orientation is taken to be the reference position for all subsequent calculations. The ever increasing rate of development of computer and image processing hardware ensures that these limitations will soon be overcome, enabling video-based systems to measure 3D eye position non-invasively with a temporal and spatial resolution comparable to the scleral search coil technique. The reference position of the eye is defined as the position where the center of the pupil lies on the hj axis. Horizontal and vertical eye position is calculated from the pupil center, which can be determined in a number of ways, such as center of mass or fitting an ellipse to the pupil boundary.

Performance of astronaut pilots during space shuttle landing was degraded after a few weeks of microgravity exposure, and longer-term exposure has the potential to impact operator proficiency during critical landing and post-landing operations for exploration-class missions. Full-motion simulations of operationally-relevant tasks were utilized to assess the impact of long-duration spaceflight on operator proficiency in a group of 8 astronauts assigned to the International Space Station, as well as a battery of cognitive/sensorimotor tests to determine the underlying cause of any post-flight performance decrements. A ground control group (N = 12) and a sleep restriction cohort (N = 9) were also tested to control for non-spaceflight factors such as lack of practice between pre- and post-flight testing and fatigue. On the day of return after 6 months aboard the space station, astronauts exhibited significant deficits in manual dexterity, dual-tasking and motion perception, and a striking degradation in the ability to operate a vehicle. These deficits were not primarily due to fatigue; performance on the same tasks was unaffected after a 30-h period of sleep restriction. Astronauts experienced a general post-flight malaise in motor function and motion perception, and a lack of cognitive reserve apparent only when faced with dual tasks, which had recovered to baseline by four days after landing.

The aim of the study was to bring together a combination of stationary (Radio Frequency IDentification (RFID), water flow meter) and animal-attached (accelerometer) sensors in an automated approach to record beef cattle drinking behaviour and herd water intake in grazing systems. An experiment was conducted to collect and validate data from the behaviour monitoring system. A water trough located in an enclosed water point was equipped with a water flow meter. The water point entry and exit gates were each fitted with a RFID panel reader. The eight beef heifers that grazed the experimental site wore a RFID ear tag in the right ear and a motion sensing neck collar that contained a triaxial accelerometer. The heifers had ad libitum access to the water point at all times. Sensor data and video observations were recorded over four consecutive weeks. When operational, the RFID readers correctly recorded 95% (94/99) of heifer movements in and out of the water point and were correlated (r = 0.99) to observed entry and exit times. Volumes of water recorded by the water meter were correlated (r = 0.99) to measured water volumes taken from the trough's inlet and from water in the trough while under the control of a float valve. An algorithm was developed to classify drinking using accelerometer measures of head-neck position, activity and movement frequency. The accelerometer algorithm detected 94% (98/104) of drinking events that were greater than 10 s in duration (F1 score = 77%) and was correlated (r = 0.84) to the observed duration of drinking events. Differences between observed and predicted estimates of the number of drinking events that were greater than 10 s in duration (1.6 +/- 1.1 vs. 2.0 +/- 1.8, respectively) and the time spent drinking (45.8 +/- 24.1 vs. 43.1 +/- 42.8, respectively) per heifer visit to the water yard were not significant (p > 0.05). The approach is considered reliable for recording a number of behavioural measures including the number, duration and frequency of visits per animal to a water point, the number and duration of drinking events per animal visit and the time each animal spends drinking.

Freezing of gait (FoG) is common in Parkinsonian gait and strongly relates to falls. Current clinical FoG assessments are patients' self-report diaries and experts' manual video analysis. Both are subjective and yield moderate reliability. Existing detection algorithms have been predominantly designed in subject-dependent settings. In this paper, we aim to develop an automated FoG detector for subject independent. After extracting highly relevant features, we apply anomaly detection techniques to detect FoG events. Specifically, feature selection is performed using correlation and clusterability metrics. From a list of 244 feature candidates, 36 candidates were selected using saliency and robustness criteria. We develop an anomaly score detector with adaptive thresholding to identify FoG events. Then, using accuracy metrics, we reduce the feature list to seven candidates. Our novel multichannel freezing index was the most selective across all window sizes, achieving sensitivity (specificity) of 96% (79%). On the other hand, freezing index from the vertical axis was the best choice for a single input, achieving sensitivity (specificity) of 94% (84%) for ankle and 89% (94%) for back sensors. Our subject-independent method is not only significantly more accurate than those previously reported, but also uses a much smaller window (e.g., 3 s versus 7.5 s) and/or lower tolerance (e.g., 0.4 s versus 2 s).

Autonomic dysfunction is common in Chagas disease and diabetes. Patients with either condition complicated by cardiac autonomic dysfunction face increased mortality, but no clinical predictors of autonomic dysfunction exist. Pupillary light reflexes (PLRs) may identify such patients early, allowing for intensified treatment. To evaluate the significance of PLRs, adults were recruited from the outpatient endocrine, cardiology, and surgical clinics at a Bolivian teaching hospital. After testing for Chagas disease and diabetes, participants completed conventional autonomic testing (CAT) evaluating their cardiovascular responses to Valsalva, deep breathing, and orthostatic changes. PLRs were measured using specially designed goggles, then CAT and PLRs were compared as measures of autonomic dysfunction. This study analyzed 163 adults, including 96 with Chagas disease, 35 patients with diabetes, and 32 controls. PLRs were not significantly different between Chagas disease patients and controls. Patients with diabetes had longer latency to onset of pupil constriction, slower maximum constriction velocities, and smaller orthostatic ratios than nonpatients with diabetes. PLRs correlated poorly with CAT results. A PLR-based clinical risk score demonstrated a 2.27-fold increased likelihood of diabetes complicated by autonomic dysfunction compared with the combination of blood tests, CAT, and PLRs (sensitivity 87.9%, specificity 61.3%). PLRs represent a promising tool for evaluating subclinical neuropathy in patients with diabetes without symptomatic autonomic dysfunction. Pupillometry does not have a role in the evaluation of Chagas disease patients.

The information coming from the vestibular otolith organs is important for the brain when reflexively making appropriate visual and spinal corrections to maintain balance. Symptoms related to failed balance control and navigation are commonly observed in astronauts returning from space. To investigate the effect of microgravity exposure on the otoliths, we studied the otolith-mediated responses elicited by centrifugation in a group of 25 astronauts before and after 6 mo of spaceflight. Ocular counterrolling (OCR) is an otolith-driven reflex that is sensitive to head tilt with regard to gravity and tilts of the gravito-inertial acceleration vector during centrifugation. When comparing pre- and postflight OCR, we found a statistically significant decrease of the OCR response upon return. Nine days after return, the OCR was back at preflight level, indicating a full recovery. Our large study sample allows for more general physiological conclusions about the effect of prolonged microgravity on the otolith system. A deconditioned otolith system is thought to be the cause of several of the negative effects seen in returning astronauts, such as spatial disorientation and orthostatic intolerance. This knowledge should be taken into account for future long-term space missions.