Leukemia classification relies on a detailed cytomorphological examination of Bone Marrow (BM) smear. However, applying existing deep-learning methods to it is facing two significant limitations. Firstly, these require large-scale datasets with expert annotations at the cell level for good results and typically suffer from poor generalization. Secondly, they simply treat BM as multi-class task, thus failing exploit correlation among leukemia subtypes over different hierarchies. Therefore,...

Artificially making clinical decisions for patients with multi-morbidity has long been considered a thorny problem due to the complexity of disease. Drug recommendations can assist doctors in automatically providing effective and safe drug combinations conducive treatment reducing adverse reactions. However, existing recommendation works ignored two critical information. (i) Different types medical information their interrelationships patient's visit history be used construct comprehensive...

Surf during typhoon events poses severe threats to coastal infrastructure and public safety. Traditional monitoring approaches, including in situ sensors numerical simulations, face inherent limitations capturing surf impacts—sensors are constrained by point-based measurements, while simulations require intensive computational resources for real-time monitoring. Video-based offers promising potential continuous observation, yet the development of deep learning models detection remains...

The capability of generating speech with a specific type emotion is desired for many human-computer interaction applications. Cross-speaker transfer common approach to emotional when data labels from target speakers not available model training. This paper presents novel cross-speaker system named iEmoTTS. composed an encoder, prosody predictor, and timbre encoder. encoder extracts the identity respective intensity mel-spectrogram input speech. measured by posterior probability that...

Tissue segmentation is an essential task in computational pathology. However, relevant datasets for such a pixel-level classification are hard to obtain due the difficulty of annotation, bringing obstacles training deep learning-based model. Recently, contrastive learning has provided feasible solution mitigating heavy reliance models on annotation. Nevertheless, applying loss most abstract image representations, existing frameworks focus global features, therefore, less capable encoding...

Predicting drug combinations according to patients' electronic health records is an essential task in intelligent healthcare systems, which can assist clinicians ordering safe and effective prescriptions. However, existing work either missed/underutilized the important information lying molecule structure encoding or has insufficient control over Drug-Drug Interactions (DDIs) rates within predictions. To address these limitations, we propose CSEDrug, enhances DDIs controlling by leveraging...

Abstract Spatial quantification is a critical step in most computational pathology tasks, from guiding pathologists to areas of clinical interest discovering tissue phenotypes behind novel biomarkers. To circumvent the need for manual annotations, modern methods have favoured multiple-instance learning approaches that can accurately predict whole-slide image labels, albeit at expense losing their spatial awareness. We prove mathematically model using instance-level aggregation could achieve...

Electronic medical data contains biochemical, imaging, pathological information during diagnosis and treatment. The pathology report is a kind of highly liberalized unstructured textual data, which the basis gold standard cancer very important for prognosis treatment patients. application extraction technology to reports can obtain structured that be understood analyzed by computers, helping pathologists make appropriate decisions. In this work, we proposed an attention-based graph...

Constructing large-scaled medical knowledge graphs (MKGs) can significantly boost healthcare applications for surveillance, bring much attention from recent research. An essential step in constructing large-scale MKG is extracting information reports. Recently, extraction techniques have been proposed and show promising performance biomedical extraction. However, these methods only consider limited types of entity relation due to the noisy text data with complex correlations. Thus, they fail...

A mouse hybrid hybridoma (tetradoma) was derived from fusing hybridomas producing monoclonal antibody to N-methylcarbamate pesticide carbofuran with organophosphorus Triazophos. The prepared tetradoma line (12C1 2H12) secreted immunoglobulin exhibiting parental and bispecific binding characteristics. effect of relevant physicochemical factors on the immunoassay based 12C1 2H12 had been studied optimize ELISA performance. developed showed that detection limit (I(20)) were 0.36 1.89 ng/mL for...

Claims database and electronic health records do not usually capture kinship or family relationship information, which is imperative for genetic research. We identify online obituaries as a new data source propose special named entity recognition relation extraction solution to extract names kinships from obituaries. Built on 1,809 annotated novel tagging scheme, our joint neural model achieved macro-averaged precision, recall F measure of 72.69%, 78.54% 74.93%, micro-averaged 95.74%, 98.25%...

Diagnostic pathology, which is the basis and gold standard of cancer diagnosis, provides essential information on prognosis disease vital evidence for clinical treatment. However, pathological diagnosis subjective, differences in observation between pathologists are common. This phenomenon more evident hospitals with insufficient medical resources. Deep learning (DL) can be used to identify classify structures digital pathology. In order solve above difficulties, this work, we propose a DL...

Biomedical relation extraction seeks to automatically extract biomedical relations from text, which plays an important role in studies. However, constructing high-quality annotation data is not only time-consuming but also requires a high level of knowledge the field. To alleviate this problem, Semi-supervised Relation Extraction aims facts limited labeled and more readily available unlabeled samples. Existing works can be roughly categorized as self-training methods self-ensembling methods....

Pathological diagnosis is the gold standard for cancer diagnosis, but it labor-intensive, in which tasks such as cell detection, classification, a nd c ounting re particularly prominent. A common solution automating these using nucleus segmentation technology. However, hard to train robust model, due several challenging problems,i.e., adhesion, stacking, and excessive fusion with background. Recently, some researchers proposed series of automatic methods based on point annotation, can...

Digital pathology plays a crucial role in the development of artificial intelligence medical field. The digital platform can make pathological resources and networked, realize permanent storage visual data synchronous browsing processing without limitation time space. It has been widely used various fields pathology. However, there is still lack an open universal to assist doctors management analysis sections, as well structured description relevant patient information. Most platforms cannot...

In this work, we present the world's first under-display lensless facial recognition system, which consists of a transparent micro-LED display, specially designed mask for amplitude modulation, CMOS sensor, and deep learning model. By utilizing kind optical component, system can optically encrypt input information, ensuring that light field information at imaging plane is incomprehensible by humans. Compared to current technologies images, advantage approach never captures any clear...

Personalized diagnoses have not been possible due to a sear amount of data pathologists bear during the day-to-day routine, leading current generalized standards being continuously updated as new findings are reported. It is noticeable that these practical developed based on multi-source heterogeneous data, including whole-slide images and pathology clinical reports. In this study, we propose framework combines pathological medical reports generate personalized diagnosis result for an...

In this work, we present the world's first under-display lensless facial recognition system, which consists of a transparent micro-LED display, specially designed mask for amplitude modulation, CMOS sensor, and deep learning model. By utilizing kind optical component, system can optically encrypt input information, ensuring that light field information at imaging plane is incomprehensible by humans. Compared to current technologies images, advantage approach never captures any clear...