Multi-class pest detection is one of the crucial components in pest management involving localization in addition to classification which is much more difficult than generic object detection because of the apparent differences among pest species. This paper proposes a region-based end-to-end approach named PestNet for large-scale multi-class pest detection and classification based on deep learning. PestNet consists of three major parts. First, a novel module channel-spatial attention (CSA) is proposed to be fused into the convolutional neural network (CNN) backbone for feature extraction and enhancement. The second one is called region proposal network (RPN) that is adopted for providing region proposals as potential pest positions based on extracted feature maps from images. Position-sensitive score map (PSSM), the third component, is used to replace fully connected (FC) layers for pest classification and bounding box regression. Furthermore, we apply contextual regions of interest (RoIs) as contextual information of pest features to improve detection accuracy. We evaluate PestNet on our newly collected large-scale pests' image dataset, Multi-class Pests Dataset 2018 (MPD2018) captured by our designed task-specific image acquisition equipment, covering more than 80k images with over 580k pests labeled by agricultural experts and categorized in 16 classes. The experimental results show that the proposed PestNet performs well on multi-class pest detection with 75.46% mean average precision (mAP), which outperforms the state-of-theart methods. INDEX TERMS Channel-spatial attention, convolutional neural network, multi-class pest detection, position-sensitive score map, region proposal network.

We propose a methodology to aid clinicians in performing lumbar spinal stenosis detection through semantic segmentation and delineation of magnetic resonance imaging (MRI) scans of the lumbar spine using deep learning. Our dataset contains MRI studies of 515 patients with symptomatic back pains. Each study is annotated by expert radiologists with notes regarding the observed characteristics and condition of the lumbar spine. We have developed a ground truth dataset, containing image labels of four important regions in the lumbar spine, to be used as the training and test images to develop classification models for segmentation. We developed two novel metrics, namely confidence, and consistency, to assess the quality of the ground truth dataset through a derivation of the Jaccard Index. We experimented with semantic segmentation of our dataset using SegNet. Our evaluation of the segmentation and the delineation results show that our proposed methodology produces a very good performance as measured by several contourbased and region-based metrics. In addition, using the Cohen's kappa and frequency-weighted confidence metrics, we can show that 1) the model's performance is within the range of the worst and the best manual labeling results and 2) the ground-truth dataset has an excellent inter-rater agreement score. We also presented two representative delineation results of the worst and best segmentation based on their BF-score to show visually how accurate and suitable the results are for computer-aided-diagnosis purposes.

No abstract

No abstract

Lumbar Spinal Stenosis causes low back pain through pressures exerted on the spinal nerves. This can be verified by measuring the anteroposterior diameter and foraminal widths of the patient’s lumbar spine. Our goal is to develop a novel strategy for assessing the extent of Lumbar Spinal Stenosis by automatically calculating these distances from the patient’s lumbar spine MRI. Our method starts with a semantic segmentation of T1- and T2-weighted composite axial MRI images using SegNet that partitions the image into six regions of interest. They consist of three main regions-of-interest, namely the Intervertebral Disc, Posterior Element, and Thecal Sac, and three auxiliary regions-of-interest that includes the Area between Anterior and Posterior elements. A novel contour evolution algorithm is then applied to improve the accuracy of the segmentation results along important region boundaries. Nine anatomical landmarks on the image are located by delineating the region boundaries found in the segmented image before the anteroposterior diameter and foraminal widths can be measured. The performance of the proposed algorithm was evaluated through a set of experiments on the Lumbar Spine MRI dataset containing MRI studies of 515 patients. These experiments compare the performance of our contour evolution algorithm with the Geodesic Active Contour and Chan-Vese methods over 22 different setups. We found that our method works best when our contour evolution algorithm is applied to improve the accuracy of both the label images used to train the SegNet model and the automatically segmented image. The average error of the calculated right and left foraminal distances relative to their expert-measured distances are 0.28 mm ( p = 0.92) and 0.29 mm ( p = 0.97), respectively. The average error of the calculated anteroposterior diameter relative to their expert-measured diameter is 0.90 mm ( p = 0.92). The method also achieves 96.7% agreement with an expert opinion on determining the severity of the Intervertebral Disc herniations.