Improved PAA algorithm for breast mass detection in mammograms.

Mammography screening is instrumental in the early detection and diagnosis of breast cancer by identifying masses in mammograms. With the rapid development of deep learning, numerous deep learning-based object detection algorithms have been explored for mass detection studies. However, these methods often yield a high false positive rate per image (FPPI) while achieving a high true positive rate (TPR). To maintain a higher TPR while also ensuring lower FPPI, we improved the Probability Anchor Assignment (PAA) algorithm to enhance the detection capability for mammographic characteristics with our previous work. We considered three dimensions: the backbone network, feature fusion module, and dense detection heads. The final experiment showed the effectiveness of the proposed method, and the TPR/FPPI values of the final improved PAA algorithm were 0.96/0.56 on the INbreast datasets. Compared to other methods, our method stands distinguished with its effectiveness in addressing the imbalance between positive and negative classes in cases of single lesion detection.

Unmanned aerial vehicle based intelligent triage system in mass-casualty incidents using 5G and artificial intelligence.

BACKGROUND: Rapid on-site triage is critical after mass-casualty incidents (MCIs) and other mass injury events. Unmanned aerial vehicles (UAVs) have been used in MCIs to search and rescue wounded individuals, but they mainly depend on the UAV operator's experience. We used UAVs and artificial intelligence (AI) to provide a new technique for the triage of MCIs and more efficient solutions for emergency rescue. METHODS: This was a preliminary experimental study. We developed an intelligent triage system based on two AI algorithms, namely OpenPose and YOLO. Volunteers were recruited to simulate the MCI scene and triage, combined with UAV and Fifth Generation (5G) Mobile Communication Technology real-time transmission technique, to achieve triage in the simulated MCI scene. RESULTS: Seven postures were designed and recognized to achieve brief but meaningful triage in MCIs. Eight volunteers participated in the MCI simulation scenario. The results of simulation scenarios showed that the proposed method was feasible in tasks of triage for MCIs. CONCLUSION: The proposed technique may provide an alternative technique for the triage of MCIs and is an innovative method in emergency rescue.

Reduced platelet 5-HT content is associated with rest tremor in Parkinson's disease.

INTRODUCTION: Parkinson's disease (PD) is highly heterogeneous in manifestations and pathogenesis. Serotonergic neurotransmitter system dysfunction is frequently implicated in PD tremor. Serotonin (5-HT) content in platelets is highly correlated with that in cerebrospinal fluid. In this study, we aimed to understand whether and how platelet 5-HT content reflects tremor in PD. METHOD: A total of 139 Chinese PD patients met with inclusion criteria were recruited. Motor and non-motor scores, and disease severity were evaluated. Patients were classified into subtypes of tremor-dominant (TD) and non-tremor-dominant (NTD). Peripheral platelets were isolated, and platelet 5-HT levels were measured. RESULTS: Platelet 5-HT content was lower in PD patients of TD subtype than in NTD subtype. Multifactor risk analysis showed that this lower content was independently associated with the TD phenotype. Platelet 5-HT level was inversely correlated with total tremor score, rest tremor amplitude score, rest tremor constancy score, and index of rest tremor, but not with postural tremor score, and kinetic tremor score. CONCLUSION: The cross-sectional study demonstrates that reduced platelet 5-HT content is associated with PD rest tremor. Our results support the involvement of serotonergic disturbance in PD rest tremor and indicate that 5-HT reduction can be manifested in peripheral platelets.

Gene-wide significant association analyses of DNMT1 genetic variants with Parkinson's disease.

Background: DNA methylation plays an important role in Parkinson's disease (PD) pathogenesis. DNA methyltransferase 1 (DNMT1) is critical for maintaining DNA methylation in mammals. The link between DNMT1 polymorphisms and PD remains elusive. Methods: The DNMT1 gene contained a total of 28 single nucleotide polymorphisms (SNPs). Four representing tag-SNPs (rs16999593, rs2162560, rs11880553, and rs9305012) were identified and genotyped in a Han Chinese population comprising 712 PD patients and 696 controls. Association analyses were performed at gene-wide significance (p < 1.8 × 10-3). Results: Rs9305012, but not the other 3 tag-SNPs, was gene-wide significantly associated with PD risk (p = 0.8 × 10-3). The rs9305012/C was a protective allele against PD (p = 1.5 × 10-3, OR 0.786, 95% CI 0.677-0.912). No significant association was observed in individual genders or PD subtypes. Haplotypes of the 4 tag-SNPs showed a significant overall distribution difference between PD patients and controls (p < 1 × 10-4). The 3-allele ACC module in the order of rs2162560, rs11880553, and rs9305012 was the highest-risk haplotype associated with PD (p < 1 × 10-4, OR 2.439, 95% CI 1.563-3.704). Rs9305012 displayed certain probability to affect transcription factor binding and target gene expression based on functional annotation analyses. Conclusion: The DNMT1 variant rs9305012 together with its haplotypes may gene-wide significantly modulate PD susceptibility. Our results support a role of DNMT1 in PD pathogenesis and provide novel insights into the genetic connection in between.

Shape Prior Guided Instance Disparity Estimation for 3D Object Detection.

In this paper, we propose a novel system named Disp R-CNN for 3D object detection from stereo images. Many recent works solve this problem by first recovering point clouds with disparity estimation and then apply a 3D detector. The disparity map is computed for the entire image, which is costly and fails to leverage category-specific prior. In contrast, we design an instance disparity estimation network (iDispNet) that predicts disparity only for pixels on objects of interest and learns a category-specific shape prior for more accurate disparity estimation. To address the challenge from scarcity of disparity annotation in training, we propose to use a statistical shape model to generate dense disparity pseudo-ground-truth without the need of LiDAR point clouds, which makes our system more widely applicable. Experiments on the KITTI dataset show that, when LiDAR ground-truth is not used at training time, Disp R-CNN outperforms previous state-of-the-art methods based on stereo input by 20 percent in terms of average precision for all categories. The code and pseudo-ground-truth data are available at the project page: https://github.com/zju3dv/disprcnn.