Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

Trajectory-level fog detection based on in-vehicle video camera with TensorFlow deep learning utilizing SHRP2 naturalistic driving data.

In Accident; analysis and prevention

Providing drivers with real-time weather information and driving assistance during adverse weather, including fog, is crucial for safe driving. The primary focus of this study was to develop an affordable in-vehicle fog detection method, which will provide accurate trajectory-level weather information in real-time. The study used the SHRP2 Naturalistic Driving Study (NDS) video data and utilized several promising Deep Learning techniques, including Deep Neural Network (DNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN). Python programming on the TensorFlow Machine Learning library has been used for training the Deep Learning models. The analysis was done on a dataset consisted of three weather conditions, including clear, distant fog and near fog. During the training process, two optimizers, including Adam and Gradient Descent, have been used. While the overall prediction accuracy of the DNN, RNN, LSTM, and CNN using the Gradient Descent optimizer were found to be around 85 %, 77 %, 84 %, and 97 %, respectively; much improved overall prediction accuracy of 88 %, 91 %, 93 %, and 98 % for the DNN, RNN, LSTM, and CNN, respectively, were observed considering the Adam optimizer. The proposed fog detection method requires only a single video camera to detect weather conditions, and therefore, can be an inexpensive option to be fitted in maintenance vehicles to collect trajectory-level weather information in real-time for expanding as well as updating weather-based Variable Speed Limit (VSL) systems and Advanced Traveler Information Systems (ATIS).

Khan Md Nasim, Ahmed Mohamed M


Advanced Driver Assistance Systems, Advanced Travel Information Systems, Convolutional Neural Network, Deep Learning, Deep Neural Network, Foggy weather, Image Classification, Long Short-Term Memory, Machine Learning, Mobile Weather Sensors, Recurrent Neural Network, TensorFlow, Variable Speed Limit

General General

Direct determination of aberration functions in microscopy by an artificial neural network.

In Optics express

Adaptive optics relies on the fast and accurate determination of aberrations but is often hindered by wavefront sensor limitations or lengthy optimization algorithms. Deep learning by artificial neural networks has recently been shown to provide determination of aberration coefficients from various microscope metrics. Here we numerically investigate the direct determination of aberration functions in the pupil plane of a high numerical aperture microscope using an artificial neural network. We show that an aberration function can be determined from fluorescent guide stars and used to improve the Strehl ratio without the need for reconstruction from Zernike polynomial coefficients.

Cumming Benjamin P, Gu Min


General General

Detecting lane change maneuvers using SHRP2 naturalistic driving data: A comparative study machine learning techniques.

In Accident; analysis and prevention

Lane change has been recognized as a challenging driving maneuver and a significant component of traffic safety research. Developing a real-time continuous lane change detection system can assist drivers to perform and deal with complex driving tasks or provide assistance when it is needed the most. This study proposed trajectory-level lane change detection models based on features from vehicle kinematics, machine vision, roadway characteristics, and driver demographics under different weather conditions. To develop the models, the SHRP2 Naturalistic Driving Study (NDS) and Roadway Information Database (RID) datasets were utilized. Initially, descriptive statistics were utilized to investigate the lane change behavior, which revealed significant differences among different weather conditions for most of the parameters. Six data fusion categories were introduced for the first time, considering different data availability. In order to select relevant features in each category, Boruta, a wrapper-based algorithm was employed. The lane change detection models were trained, validated, and comparatively evaluated using four Machine Learning algorithms including Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and eXtrem Gradient Boosting (XGBoost). The results revealed that the highest overall detection accuracy was found to be 95.9 % using the XGBoost model when all the features were included in the model. Moreover, the highest overall detection accuracy of 81.9 % using the RF model was achieved considering only vehicle kinematics-based features, indicating that the proposed model could be utilized when other data are not available. Furthermore, the analysis of the impact of weather conditions on lane change detection suggested that incorporating weather could improve the accuracy of lane change detection. In addition, the analysis of early lane change detection indicated that the proposed algorithm could predict the lane changes within 5 s before the vehicles cross the lane line. The developed detection models could be used to monitor and control driver behavior in a Cooperative Automated Vehicle environment.

Das Anik, Khan Md Nasim, Ahmed Mohamed M


Artificial neural network, Connected vehicle, Lane change detection, Naturalistic driving study, Random Forest, Support vector machine, eXtrem gradient boosting

General General

The emergence of cardiac changes following the self-administration of methamphetamine.

In Drug and alcohol dependence ; h5-index 64.0

BACKGROUND : Clinical observations suggest an association between methamphetamine (METH) use and cardiovascular disease, but preclinical studies are lacking. The purpose of the current study was to explore changes in left ventricular function as a potential precursor to cardiovascular disease in a rodent model of METH use.

METHODS : Male rats were allowed to self-administer either METH or saline for 9 d. On the day following the 4th and 9th self-administration sessions, an echocardiogram was performed to assess left-ventricular parameters under basal conditions and following a low-dose of METH (1 mg/kg).

RESULTS : A low challenge dose of METH resulted in subtle but statistically significant changes in cardiac function during the echocardiogram in both the METH and saline self-administering groups. Further, differences in left-ventricular parameters such as stroke volume and heart rate were observed between METH and saline groups following the 9th self-administration session. Finally, supervised machine learning correctly predicted the self-administration group assignment (saline or METH) using cardiac parameters following the 9th self-administration session.

CONCLUSIONS : The findings of the current study suggest the heart, specifically the left ventricle, is sensitive to METH. Overall, these findings and emerging clinical observations highlight the need for research to investigate the effects of METH use on the heart.

Freeling Jessica L, McFadden Lisa M


Echocardiogram, Heart, Left ventricle, Methamphetamine, Self-administration

Radiology Radiology

Radiological classification of dementia from anatomical MRI assisted by machine learning-derived maps.

In Journal of neuroradiology = Journal de neuroradiologie

BACKGROUND AND PURPOSE : Many artificial intelligence tools are currently being developed to assist diagnosis of dementia from magnetic resonance imaging (MRI). However, these tools have so far been difficult to integrate in the clinical routine workflow. In this work, we propose a new simple way to use them and assess their utility for improving diagnostic accuracy.

MATERIALS AND METHODS : We studied 34 patients with early-onset Alzheimer's disease (EOAD), 49 with late-onset AD (LOAD), 39 with frontotemporal dementia (FTD) and 24 with depression from the pre-existing cohort CLIN-AD. Support vector machine (SVM) automatic classifiers using 3D T1 MRI were trained to distinguish: LOAD vs Depression, FTD vs LOAD, EOAD vs Depression, EOAD vs FTD. We extracted SVM weight maps, which are tridimensional representations of discriminant atrophy patterns used by the classifier to take its decisions and we printed posters of these maps. Four radiologists (2 senior neuroradiologists and 2 unspecialized junior radiologists) performed a visual classification of the 4 diagnostic pairs using 3D T1 MRI. Classifications were performed twice: first with standard radiological reading and then using SVM weight maps as a guide.

RESULTS : Diagnostic performance was significantly improved by the use of the weight maps for the two junior radiologists in the case of FTD vs EOAD. Improvement was over 10 points of diagnostic accuracy.

CONCLUSION : This tool can improve the diagnostic accuracy of junior radiologists and could be integrated in the clinical routine workflow.

Chagué Pierre, Marro Béatrice, Fadili Sarah, Houot Marion, Morin Alexandre, Samper-González Jorge, Beunon Paul, Arrivé Lionel, Dormont Didier, Dubois Bruno, Teichmann Marc, Epelbaum Stéphane, Colliot Olivier


“Alzheimers disease”, Dementia, anatomical MRI, artificial intelligence, diagnosis

General General

Ca2+ imaging of neurons in freely moving rats with automatic post hoc histological identification.

In Journal of neuroscience methods

BACKGROUND : Cognitive neuroscientists aim to understand behavior often based on the underlying activity of individual neurons. Recently developed miniaturized epifluorescence microscopes allow recording of cellular calcium transients, resembling neuronal activity, of individual neurons even in deep brain areas in freely behaving animals. At the same time, molecular markers allow the characterization of diverse neuronal subtypes by post hoc immunohistochemical labeling. Combining both methods would allow researchers to increase insights into how individual neuronal activity and entities contribute to behavior.

NEW METHOD : Here, we present a novel method for identifying the same neurons, recorded with calcium imaging using a miniaturized epifluorescence microscope, post hoc in fixed histological sections. This allows immunohistochemical investigations to detect the molecular signature of in vivo recorded neurons. Our method utilizes the structure of blood vessels for aligning in vivo acquired 2D images with a reconstructed 3D histological model.

RESULTS : We automatically matched, 60 % of all in vivo recorded cells post hoc in histology. Across all animals, we successfully matched 43 % to 89 % of the recorded neurons. We provide a measure for the confidence of matched cells and validated our method by multiple simulation studies.

COMPARISON WITH EXISTING METHODS : To our knowledge, we present the first method for matching cells, recorded with a miniaturized epifluorescence microscope in freely moving animals, post hoc in histological sections.

CONCLUSIONS : Our method allows a comprehensive analysis of how cortical circuits relate to freely moving animal behavior by combining functional activity of individual neurons with their underlying histological profiles.

Anner Philip, Passecker Johannes, Klausberger Thomas, Dorffner Georg


calcium imaging, confocal microscopy, freely moving behavior, histology, immunohistochemistry, miniature microscopes