Predicting incomplete occlusion of intracranial aneurysms treated with flow diverters using machine learning models.

OBJECTIVE: Intracranial saccular aneurysms are vascular malformations responsible for 80% of nontraumatic brain hemorrhage. Recently, flow diverters have been used as a less invasive therapeutic alternative for surgery. However, they fail to achieve complete occlusion after 6 months in 25% of cases. In this study, the authors built a tool, using machine learning (ML), to predict the aneurysm occlusion outcome 6 months after treatment with flow diverters. METHODS: A total of 667 aneurysms in 616 patients treated with the Pipeline embolization device at a tertiary referral center between January 2011 and December 2017 were included. To build the predictive tool, two experiments were conducted. In the first experiment, six ML algorithms (support vector machine [SVM], decision tree, random forest [RF], k-nearest neighbor, XGBoost, and CatBoost) were trained using 26 features related to patient risk factors and aneurysm morphological characteristics, and the results were compared with logistic regression (LR) modeling. In the second experiment, the models were trained using the top 10 features extracted by Shapley additive explanation (SHAP) analysis performed on the RF model. RESULTS: The results showed that the authors' tool can better predict the occlusion outcome than LR (accuracy of 89% for the SVM model vs 62% for the LR model), even when trained using a subset of the features (83% accuracy). SHAP analysis revealed that age, hypertension, smoking status, branch vessel involvement, aneurysm neck, and larger diameter dimensions were among the most important features contributing to accurate predictions. CONCLUSIONS: In this study, an ML-based tool was developed that successfully predicts outcome in intracranial aneurysms treated with flow diversion, thus helping neurosurgeons to practice a more refined approach and patient-tailored medicine.

GAN-based semi-automated augmentation online tool for agricultural pest detection: A case study on whiteflies.

Deep neural networks can be used to diagnose and detect plant diseases, helping to avoid the plant health-related crop production losses ranging from 20 to 50% annually. However, the data collection and annotation required to achieve high accuracies can be expensive and sometimes very difficult to obtain in specific use-cases. To this end, this work proposes a synthetic data generation pipeline based on generative adversarial networks (GANs), allowing users to artificially generate images to augment their small datasets through its web interface. The image-generation pipeline is tested on a home-collected dataset of whitefly pests, Bemisia tabaci, on different crop types. The data augmentation is shown to improve the performance of lightweight object detection models when the dataset size is increased from 140 to 560 images, seeing a jump in recall at 0.50 IoU from 54.4 to 93.2%, and an increase in the average IoU from 34.6 to 70.9%, without the use of GANs. When GANs are used to increase the number of source object masks and further diversify the dataset, there is an additional 1.4 and 2.6% increase in recall and average IoU, respectively. The authenticity of the generated data is also validated by human reviewers, who reviewed the GANs generated data and scored an average of 56% in distinguishing fake from real insects for low-resolutions sets, and 67% for high-resolution sets.

Acceptability of smokers of a conceptual cigarette tracker as wearable for smoking reduction.

OBJECTIVE: The study aims to explore smokers' acceptance of using a conceptual cigarette tracker like a cigarette filter for smoking cessation using the Technology Acceptance Model (TAM). Smokers presenting to the family medicine clinics at a tertiary care center were asked to complete an anonymous questionnaire. RESULTS: A total of 45 participants were included. Two-thirds of the smokers reported that they would like to try such a tracker and perceived its usefulness in reducing the number of daily cigarettes consumed and increasing the motivation to join a smoking cessation program. A range of 40-50% of the participants had a neutral attitude towards the visibility of the tracker and its effect on social acceptance and self-image. The structural equation model with latent variables path analysis showed that only perceived usefulness correlated to the intention to adopt with statistical significance. Visibility was correlated with intention to adopt with a marginal p-value of 0.061. Driven by perceived usefulness, smokers may buy or try a cigarette tracker for smoking reduction or cessation.

The power of transfer learning in agricultural applications: AgriNet.

Advances in deep learning and transfer learning have paved the way for various automation classification tasks in agriculture, including plant diseases, pests, weeds, and plant species detection. However, agriculture automation still faces various challenges, such as the limited size of datasets and the absence of plant-domain-specific pretrained models. Domain specific pretrained models have shown state of art performance in various computer vision tasks including face recognition and medical imaging diagnosis. In this paper, we propose AgriNet dataset, a collection of 160k agricultural images from more than 19 geographical locations, several images captioning devices, and more than 423 classes of plant species and diseases. We also introduce AgriNet models, a set of pretrained models on five ImageNet architectures: VGG16, VGG19, Inception-v3, InceptionResNet-v2, and Xception. AgriNet-VGG19 achieved the highest classification accuracy of 94% and the highest F1-score of 92%. Additionally, all proposed models were found to accurately classify the 423 classes of plant species, diseases, pests, and weeds with a minimum accuracy of 87% for the Inception-v3 model. Finally, experiments to evaluate of superiority of AgriNet models compared to ImageNet models were conducted on two external datasets: pest and plant diseases dataset from Bangladesh and a plant diseases dataset from Kashmir.

TempoMAGE: a deep learning framework that exploits the causal dependency between time-series data to predict histone marks in open chromatin regions at time-points with missing ChIP-seq datasets.

MOTIVATION: Identifying histone tail modifications using ChIP-seq is commonly used in time-series experiments in development and disease. These assays, however, cover specific time-points leaving intermediate or early stages with missing information. Although several machine learning methods were developed to predict histone marks, none exploited the dependence that exists in time-series experiments between data generated at specific time-points to extrapolate these findings to time-points where data cannot be generated for lack or scarcity of materials (i.e. early developmental stages). RESULTS: Here, we train a deep learning model named TempoMAGE, to predict the presence or absence of H3K27ac in open chromatin regions by integrating information from sequence, gene expression, chromatin accessibility and the estimated change in H3K27ac state from a reference time-point. We show that adding reference time-point information systematically improves the overall model's performance. In addition, sequence signatures extracted from our method were exclusive to the training dataset indicating that our model learned data-specific features. As an application, TempoMAGE was able to predict the activity of enhancers from pre-validated in-vivo dataset highlighting its ability to be used for functional annotation of putative enhancers. AVAILABILITY AND IMPLEMENTATION: TempoMAGE is freely available through GitHub at https://github.com/pkhoueiry/TempoMAGE. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Anytime multipurpose emotion recognition from EEG data using a Liquid State Machine based framework.

Recent technological advances in machine learning offer the possibility of decoding complex datasets and discern latent patterns. In this study, we adopt Liquid State Machines (LSM) to recognize the emotional state of an individual based on EEG data. LSM were applied to a previously validated EEG dataset where subjects view a battery of emotional film clips and then rate their degree of emotion during each film based on valence, arousal, and liking levels. We introduce LSM as a model for an automatic feature extraction and prediction from raw EEG with potential extension to a wider range of applications. We also elaborate on how to exploit the separation property in LSM to build a multipurpose and anytime recognition framework, where we used one trained model to predict valence, arousal and liking levels at different durations of the input. Our simulations showed that the LSM-based framework achieve outstanding results in comparison with other works using different emotion prediction scenarios with cross validation.

Order reduction and efficient implementation of nonlinear nonlocal cochlear response models.

The cochlea is an indispensable preliminary processing stage in auditory perception that employs mechanical frequency-tuning and electrical transduction of incoming sound waves. Cochlear mechanical responses are shown to exhibit active nonlinear spatiotemporal response dynamics (e.g., otoacoustic emission). To model such phenomena, it is often necessary to incorporate cochlear fluid-membrane interactions. This results in both excessively high-order model formulations and computationally intensive solutions that limit their practical use in simulating the model and analyzing its response even for simple single-tone inputs. In order to address these limitations, the current work employs a control-theoretic framework to reformulate a nonlinear two-dimensional cochlear model into discrete state space models that are of considerably lower order (factor of 8) and are computationally much simpler (factor of 25). It is shown that the reformulated models enjoy sparse matrix structures which permit efficient numerical manipulations. Furthermore, the spatially discretized models are linearized and simplified using balanced transformation techniques to result in lower-order (nonlinear) realizations derived from the dominant Hankel singular values of the system dynamics. Accuracy and efficiency of the reduced-order reformulations are demonstrated under the response to two fixed tones, sweeping tones and, more generally, a brief speech signal. The corresponding responses are compared to those produced by the original model in both frequency and spatiotemporal domains. Although carried out on a specific instance of cochlear models, the introduced framework of control-theoretic model reduction could be applied to a wide class of models that address the micro- and macro-mechanical properties of the cochlea.