Fisher's Linear Discriminant Analysis With Space-Folding Operations.

Fisher's linear discriminant analysis (LDA) is an easy-to-use supervised dimensionality reduction method. However, LDA may be ineffective against complicated class distributions. It is well-known that deep feedforward neural networks with rectified linear units as activation functions can map many input neighborhoods to similar outputs by a succession of space-folding operations. This short paper shows that the space-folding operation can reveal to LDA classification information in the subspace where LDA cannot find any. A composition of LDA with the space-folding operation can find classification information more than LDA can do. End-to-end fine-tuning can improve that composition further. Experimental results on artificial and open data sets have shown the feasibility of the proposed approach.

A Two-Mode Underwater Smart Sensor Object for Precision Aquaculture Based on AIoT Technology.

Monitoring the status of culture fish is an essential task for precision aquaculture using a smart underwater imaging device as a non-intrusive way of sensing to monitor freely swimming fish even in turbid or low-ambient-light waters. This paper developed a two-mode underwater surveillance camera system consisting of a sonar imaging device and a stereo camera. The sonar imaging device has two cloud-based Artificial Intelligence (AI) functions that estimate the quantity and the distribution of the length and weight of fish in a crowded fish school. Because sonar images can be noisy and fish instances of an overcrowded fish school are often overlapped, machine learning technologies, such as Mask R-CNN, Gaussian mixture models, convolutional neural networks, and semantic segmentation networks were employed to address the difficulty in the analysis of fish in sonar images. Furthermore, the sonar and stereo RGB images were aligned in the 3D space, offering an additional AI function for fish annotation based on RGB images. The proposed two-mode surveillance camera was tested to collect data from aquaculture tanks and off-shore net cages using a cloud-based AIoT system. The accuracy of the proposed AI functions based on human-annotated fish metric data sets were tested to verify the feasibility and suitability of the smart camera for the estimation of remote underwater fish metrics.

Homophilic ATP1A1 binding induces activin A secretion to promote EMT of tumor cells and myofibroblast activation.

Tumor cells with diverse phenotypes and biological behaviors are influenced by stromal cells through secretory factors or direct cell-cell contact. Pancreatic ductal adenocarcinoma (PDAC) is characterized by extensive desmoplasia with fibroblasts as the major cell type. In the present study, we observe enrichment of myofibroblasts in a juxta-tumoral position with tumor cells undergoing epithelial-mesenchymal transition (EMT) that facilitates invasion and correlates with a worse clinical prognosis in PDAC patients. Direct cell-cell contacts forming heterocellular aggregates between fibroblasts and tumor cells are detected in primary pancreatic tumors and circulating tumor microemboli (CTM). Mechanistically, ATP1A1 overexpressed in tumor cells binds to and reorganizes ATP1A1 of fibroblasts that induces calcium oscillations, NF-κB activation, and activin A secretion. Silencing ATP1A1 expression or neutralizing activin A secretion suppress tumor invasion and colonization. Taken together, these results elucidate the direct interplay between tumor cells and bound fibroblasts in PDAC progression, thereby providing potential therapeutic opportunities for inhibiting metastasis by interfering with these cell-cell interactions.

Fish Segmentation in Sonar Images by Mask R-CNN on Feature Maps of Conditional Random Fields.

Imaging sonar systems are widely used for monitoring fish behavior in turbid or low ambient light waters. For analyzing fish behavior in sonar images, fish segmentation is often required. In this paper, Mask R-CNN is adopted for segmenting fish in sonar images. Sonar images acquired from different shallow waters can be quite different in the contrast between fish and the background. That difference can make Mask R-CNN trained on examples collected from one fish farm ineffective to fish segmentation for the other fish farms. In this paper, a preprocessing convolutional neural network (PreCNN) is proposed to provide "standardized" feature maps for Mask R-CNN and to ease applying Mask R-CNN trained for one fish farm to the others. PreCNN aims at decoupling learning of fish instances from learning of fish-cultured environments. PreCNN is a semantic segmentation network and integrated with conditional random fields. PreCNN can utilize successive sonar images and can be trained by semi-supervised learning to make use of unlabeled information. Experimental results have shown that Mask R-CNN on the output of PreCNN is more accurate than Mask R-CNN directly on sonar images. Applying Mask R-CNN plus PreCNN trained for one fish farm to new fish farms is also more effective.

An intelligent indoor positioning system based on pedestrian directional signage object detection: a case study of Taipei Main Station.

Indoor positioning technologies have gained great interest from both industry and academia. Variety of services and applications can be built based on the availability and accessibility of indoor positioning information, for example indoor navigation and various location-based services. Different approaches have been proposed to provide indoor positioning information to users, in which an underlying system infrastructure is usually assumed to be well deployed in advance to provide the position information to users. Among many others, one common strategy is to deploy a bunch of active sensor nodes, such as WiFi APs and Bluetooth transceivers, to the indoor environment to serve as reference landmarks. The user's current location can thus be obtained directly or indirectly according to the active sensor signals collected by the user. Different from conventional infrastructure-based approaches, which put additional sensor devices to the environment, we utilize available objects in the environment as location landmarks. Leveraging wildly available smartphone devices as customer premises equipment to the user and the cutting-edge deep-learning technology, we investigate the feasibility of an infrastructure-free intelligent indoor positioning system based on visual information only. The proposed scheme has been verified by a real case study, which is to provide indoor positioning information to users in Taipei Main Station, one of the busiest transportation stations in the world. We use available pedestrian directional signage as location landmarks, which include all of the 52 pedestrian directional signs in the testing area. The Google Objection Detection framework is applied for detection and recognition of the pedestrian directional sign. According to the experimental results, we have shown that the proposed scheme can achieve as high as 98% accuracy to successfully identify the 52 pedestrian directional signs for the three test data sets which include 6,341 test images totally. Detailed discussions of the system design and the experiments are also presented in the paper.

Automatic Tuning of the RBF Kernel Parameter for Batch-Mode Active Learning Algorithms: A Scalable Framework.

Batch-mode active learning algorithms can select a batch of valuable unlabeled samples to manually annotate for reducing the total cost of labeling every unlabeled sample. To facilitate selection of valuable unlabeled samples, many batch-mode active learning algorithms map samples to the reproducing kernel Hilbert space induced by a radial-basis function (RBF) kernel. Setting a proper value to the parameter for the RBF kernel is crucial for such batch-mode active learning algorithms. In this paper, for automatic tuning of the kernel parameter, a hypothesis-margin-based criterion function is proposed. Three frameworks are also developed to incorporate the function of automatic tuning of the kernel parameter with existing batch-model active learning algorithms. In the proposed frameworks, the kernel parameter can be tuned in a single stage or in multiple stages. Tuning the kernel parameter in a single stage aims for the kernel parameter to be suitable for selecting the specified number of unlabeled samples. When the kernel parameter is tuned in multiple stages, the incorporated active learning algorithm can be enforced to make coarse-to-fine evaluations of the importance of unlabeled samples. The proposed framework can also improve the scalability of existing batch-mode active learning algorithms satisfying a decomposition property. Experimental results on data sets comprising hundreds to hundreds of thousands of samples have shown the feasibility of the proposed framework.

A Novel GMM-Based Behavioral Modeling Approach for Smartwatch-Based Driver Authentication.

All drivers have their own distinct driving habits, and usually hold and operate the steering wheel differently in different driving scenarios. In this study, we proposed a novel Gaussian mixture model (GMM)-based method that can improve the traditional GMM in modeling driving behavior. This new method can be applied to build a better driver authentication system based on the accelerometer and orientation sensor of a smartwatch. To demonstrate the feasibility of the proposed method, we created an experimental system that analyzes driving behavior using the built-in sensors of a smartwatch. The experimental results for driver authentication-an equal error rate (EER) of 4.62% in the simulated environment and an EER of 7.86% in the real-traffic environment-confirm the feasibility of this approach.

Cellular 5-hydroxylmethylcytosine content determines tumorigenic potential and prognosis of pancreatic ductal adenocarcinoma.

We stratified pancreatic ductal adenocarcinoma (PDAC) based on the tumorigenic properties of cancer cells, and aimed to identify clinically useful immunohistochemical (IHC) markers with mechanistic insights. The tumorigenic properties of PDACs were determined using patient-derived xenograft in NOD/SCID/IL2Rγnull mice. The success of tumor engraftment was significantly correlated to poor survival, and its predictive values were superior to clinicopathological parameters. To search IHC-based biomarkers as surrogate for high tumorigenicity with prognostic values, 11 candidates of potentially clinical useful prognostic markers were selected. Among them, 5hmC content of the cancer cells was validated. Elevated 5hmC content positively correlated with in vivo tumorigenicity and poor prognosis in both primary and validation cohorts. Enrichment of cancer-associated 5hmC in CDX2 and FOXA1 lineage-specific transcriptional factor genes further pointed out the potential role of 5hmC in modulating cellular differentiation to enhance tumor malignancy during PDAC progression. Tumor-associated 5hmC content defined a subpopulation of PDAC with high lineage plasticity and tumorigenic potential, and was a prognostic IHC marker that provided a clinical basis for future management of PDAC.

Active learning for semi-supervised clustering based on locally linear propagation reconstruction.

The success of semi-supervised clustering relies on the effectiveness of side information. To get effective side information, a new active learner learning pairwise constraints known as must-link and cannot-link constraints is proposed in this paper. Three novel techniques are developed for learning effective pairwise constraints. The first technique is used to identify samples less important to cluster structures. This technique makes use of a kernel version of locally linear embedding for manifold learning. Samples neither important to locally linear propagation reconstructions of other samples nor on flat patches in the learned manifold are regarded as unimportant samples. The second is a novel criterion for query selection. This criterion considers not only the importance of a sample to expanding the space coverage of the learned samples but also the expected number of queries needed to learn the sample. To facilitate semi-supervised clustering, the third technique yields inferred must-links for passing information about flat patches in the learned manifold to semi-supervised clustering algorithms. Experimental results have shown that the learned pairwise constraints can capture the underlying cluster structures and proven the feasibility of the proposed approach.

Catastrophic hemophagocytic lymphohistiocytosis in a young man with nephrotic syndrome.

BACKGROUND: In its early stage, necrotizing fasciitis mimics a milder cutaneous infection, such as cellulitis, and is an uncommon but potentially fatal complication of nephrotic syndrome. It may trigger an uncontrolled and catastrophic immune response, such as hemophagocytic lymphohistiocytosis (HLH). CASE REPORT: A 19-y-old man presented with steroid-resistant nephrotic syndrome and rapidly progressing Escherichia coli monomicrobial necrotizing fasciitis with bacteremia. The conditions developed one day after steroid therapy, leading to multiple organ dysfunction syndrome. A provisional diagnosis of HLH was promptly made, based upon the patient's fever, unremitting shock, marked pancytopenia, hyperferritinemia, hypofibrinogenemia, and the typical histiocytic hemophagocytosis in pleural effusion. Despite aggressive medical treatment and organ support, the patient died 8 days after transfer to our intensive care unit. Final bone marrow examination confirmed the diagnosis of HLH. CONCLUSIONS: Although nephrotic syndrome associated with E. coli infection is common, this is the first reported case of E. coli monomicrobial necrotizing fasciitis with bacteremia resulting in HLH in a patient with nephrotic syndrome.

A 10-year analysis of thyrotoxic periodic paralysis in 135 patients: focus on symptomatology and precipitants.

BACKGROUND: A comprehensive analysis has not been performed on patients with thyrotoxic periodic paralysis (TPP) characterized by acute hypokalemia and paralysis in the setting of thyrotoxicosis. PURPOSE: The aim of this study was to analyze the detailed symptomatology of thyrotoxicosis and precipitating factors for the attack in a large cohort of TPP patients. PATIENTS AND METHODS: A prospective observational study enrolled patients with TPP consecutively over 10 years at an academic medical center. Clinical features, including signs/symptoms of thyrotoxicosis and precipitating factors, were analyzed. The Wayne's index was used to assess the severity of thyrotoxicosis at presentation. Patients who agreed to receive an oral glucose-loading test after recovery were evaluated. RESULTS: Among the 135 TPP patients (male:female, 130:5), 70% of paralytic attacks occurred in the morning, especially during the seasons of summer and fall. Two-thirds of patients did not have a known family or personal history of hyperthyroidism. Only 17% of TPP patients manifested overt signs/symptoms of thyrotoxicosis (Wayne's index >19). A clear precipitating factor, such as high carbohydrate load, acute upper respiratory tract infection, strenuous exercise, high-salt diet, or the use of steroids or bronchodilators, was identified in only 34% of TPP patients. A glucose load to stimulate insulin secretion induced acute hypokalemia (K(+)2.47±0.6 mmol/l) with reparalysis in only 18% (10/55) of TPP patients. CONCLUSIONS: Most TPP patients have only subtle clinical signs/symptoms of thyrotoxicosis and only a small fraction has clear precipitating factors. In addition to the effects of hyperinsulinemia, other insulin-independent mechanisms may participate in the pathogenesis of TPP.

Uremic tumoral calcinosis causing atlantoaxial subluxation and spinal cord compression in a patient on continuous ambulatory peritoneal dialysis.

Uremic tumoral calcinosis (UTC) is a form of metastatic tissue calcification unique to dialysis patients, manifesting with amorphous and cystic masses containing calcium phosphate deposits in periarticular soft tissue. An involvement of the cervical spine with bone destruction is extremely rare in UTC. We describe a 44-year-old uremic female on long-term continuous ambulatory peritoneal dialysis who developed UTC in the peri-odontoid region with consequent atlantoaxial subluxation and spinal cord compression, featuring severe neck soreness, headache, and hypertension. Surgical removal of the destructive cervical spine lesion, showing typical tumoral calcinosis on histology, completely resolved the clinical symptoms. To date, the patient maintains uneventful postoperative course with tight control of serum phosphorus, calcium, and secondary hyperparathyroidism by medical treatment. We also review other reported unusual cases of UTC involving the cervical spine and discuss the differential diagnosis of destructive spinal lesions in uremic patients, such as UTC, dialysis-related amyloidosis, and brown tumors.

Hypokalemic paralysis in a young obese female.

BACKGROUND: Profound hypokalemia with paralysis usually poses a diagnostic and therapeutic challenge. METHODS: We report on a 28-y-old obese Chinese female presenting with sudden onset of flaccid quadriparesis upon awaking in the morning. There is no family history of hyperthyroidism. She experienced body weight loss of 7 kg in 2 months. RESULTS: The most conspicuous blood biochemistry is marked hypokalemia (1.8 mmol/l) and hypophosphatemia (0.5 mmol/l) associated with low urine K(+) and phosphate excretion. Surreptitious laxatives and/or diuretics abuse-related hypokalemic paralysis were tentatively made. However, her relatively normal blood acid-base status and the absence of low urine Na(+) and/or Cl(-) excretion made these diagnoses unlikely. Furthermore, she developed rebound hyperkalemia (5.7 mmol/l) after only 80 mmol K(+) supplementation. Thyroid function test confirmed hyperthyroidism due to Graves' disease. Control of the hyperthyroidism completely abolished her periodic paralysis. CONCLUSIONS: Thyrotoxic periodic paralysis (TPP) should be kept in mind as a cause of paralysis in female, even with obesity, despite its predominance in adult males.

Deformable shape finding with models based on kernel methods.

In this paper, a new kernel-based deformable model is proposed for detecting deformable shapes. To incorporate valuable information for shape detection, such as edge orientations into the shape representation, a novel scheme based on kernel methods has been utilized. The variation model of a deformable shape is established by a set of training samples of the shape represented in a kernel feature space. The proposed deformable model consists of two parts: a set of basis vectors describing the sample subspace, including the shape representations of the training samples, and a feasibility constraint generated by the one-class support vector machine to describe the feasible region of the training samples in the sample subspace. The aim of the proposed feasibility constraint is to avoid finding some invalid shapes. By using the proposed deformable model, an efficient algorithm without initial solutions is developed for shape detection. The proposed approach was tested against real images. Experimental results show the effectiveness of the proposed deformable model and prove the feasibility of the proposed approach.