Measurement of the liver and spleen volumes has clinical implications. Although computed tomography (CT) volumetry is considered to be most reliable noninvasive method for volume measurement, it limited application in practice due its time-consuming segmentation process. We aimed develop validate a deep learning algorithm (DLA) fully automated using portal venous phase CT images various conditions.A DLA was trained development dataset from 813 patients. Performance evaluated two separate...

In recent years, deep learning-based feature representation methods have shown a promising impact on electroencephalography (EEG)-based brain–computer interface (BCI). Nonetheless, owing to high intra- and inter-subject variabilities, many studies decoding EEG were designed in subject-specific manner by using calibration samples, with no concern of its practical use, hampered time-consuming steps large data requirement. To this end, adopted transfer learning strategy, especially domain...

Identifying Alzheimer's disease (AD) involves a deliberate diagnostic process owing to its innate traits of irreversibility with subtle and gradual progression. These characteristics make AD biomarker identification from structural brain imaging (e.g., MRI) scans quite challenging. Using clinically-guided prototype learning, we propose novel deep-learning approach through eXplainable Likelihood Map Estimation (XADLiME) for progression modeling over 3D sMRIs. Specifically, establish set...

Existing studies on disease diagnostic models focus either model learning for performance improvement or the visual explanation of a trained model. We propose novel learn-explain-reinforce (LEAR) framework that unifies learning, generation (explanation unit), and reinforcement (reinforcement unit) guided by explanation. For explanation, we generate counterfactual map transforms an input sample to be identified as intended target label. example, can localize hypothetical abnormalities within...

In this paper, we propose a novel architecture of deep neural network for EEG-based motor imagery classification. Unlike the existing networks in literature, proposed allows us to analyze learned weights from neurophysiological perspective, thus providing an insight into underlying patterns inherent induced EEG signals. order validate effectiveness method, conducted experiments on BCI Competition IV-IIa dataset by comparing with competing methods terms Cohen's k value. For qualitative...

We aimed to develop and test a deep learning algorithm (DLA) for fully automated measurement of the volume signal intensity (SI) liver spleen using gadoxetic acid-enhanced hepatobiliary phase (HBP)-magnetic resonance imaging (MRI) evaluate clinical utility DLA-assisted assessment functional capacity.The DLA was developed HBP-MRI data from 1014 patients. Using an independent dataset (110 internal 90 external MRI data), segmentation performance measured Dice similarity score (DSS), agreement...

CT volumetry (CTV) has been widely used for pre-operative graft weight (GW) estimation in living-donor liver transplantation (LDLT), and the use of a deep-learning algorithm (DLA) may further improve its efficiency. However, accuracy not well determined. To evaluate efficiency DLA-assisted CTV GW estimation, we performed retrospective study including 581 consecutive LDLT donors who donated right-lobe graft. Right-lobe volume (GV) was measured on using software implemented with DLA automated...

Background Reference intervals guiding volumetric assessment of the liver and spleen have yet to be established. Purpose To establish population-based personalized reference for volume, liver-to-spleen volume ratio (LSVR). Materials Methods This retrospective study consecutively included healthy adult donors from 2001 2013 (reference group) 2014 2016 (healthy validation patients with viral hepatitis 2007 2017. Liver LSVR were measured CT by using a deep learning algorithm. In group, indexes...

Although the liver-to-spleen volume ratio (LSVR) based on CT reflects portal hypertension, its prognostic role in cirrhotic patients has not been proven. We evaluated utility of LSVR, automatically measured from images using a deep learning algorithm, as predictor hepatic decompensation and transplantation-free survival with hepatitis B viral (HBV)-compensated cirrhosis.A algorithm was used to measure LSVR cohort 1027 consecutive (mean age, 50.5 years; 675 male 352 female) HBV-compensated...

Convolutional neural networks (CNNs), which can recognize structural/configuration patterns in data with different architectures, have been studied for feature extraction. However, challenges remain regarding leveraging advanced deep learning methods BCIs. We focus on problems of small-sized training samples and interpretability the learned parameters leverages a semi-supervised generative discriminative framework that effectively utilizes synthesized real to discover class-discriminative...

This study aimed to evaluate the usefulness of quantitative indices obtained from deep learning analysis gadoxetic acid-enhanced hepatobiliary phase (HBP) MRI and their longitudinal changes in predicting decompensation death patients with advanced chronic liver disease (ACLD).We included who underwent baseline 1-year follow-up a prospective cohort that for hepatocellular carcinoma surveillance between November 2011 August 2012 at tertiary medical center. Baseline condition was categorized as...

Medical image analysis (MedIA) has become an essential tool in medicine and healthcare, aiding disease diagnosis, prognosis, treatment planning, recent successes deep learning (DL) have made significant contributions to its advances. However, deploying DL models for MedIA real-world situations remains challenging due their failure generalize across the distributional gap between training testing samples - a problem known as domain shift. Researchers dedicated efforts developing various...

Diagnosing Alzheimer's disease (AD) involves a deliberate diagnostic process owing to its innate traits of irreversibility with subtle and gradual progression. These characteristics make AD biomarker identification from structural brain imaging (e.g., MRI) scans quite challenging. Furthermore, there is high possibility getting entangled normal aging. We propose novel deep-learning approach through eXplainable Likelihood Map Estimation (XADLiME) for progression modeling over 3D sMRIs using...

In this article, we focus on decomposing latent representations in generative adversarial networks or learned feature deep autoencoders into semantically controllable factors a semisupervised manner, without modifying the original trained models. Particularly, propose factors' decomposer-entangler network (FDEN) that learns to decompose representation mutually independent factors. Given representation, proposed framework draws set of interpretable factors, each aligned variations by...

There exists an apparent negative correlation between performance and interpretability of deep learning models. In effort to reduce this correlation, we propose a Born Identity Network (BIN), which is post-hoc approach for producing multi-way counterfactual maps. A map transforms input sample be conditioned classified as target label, similar how humans process knowledge through thinking. For example, can localize hypothetical abnormalities from normal brain image that may cause it diagnosed...

In recent years, deep learning-based feature representation methods have shown a promising impact in electroencephalography (EEG)-based brain-computer interface (BCI). Nonetheless, owing to high intra- and inter-subject variabilities, many studies on decoding EEG were designed subject-specific manner by using calibration samples, with no concern of its practical use, hampered time-consuming steps large data requirement. To this end, adopted transfer learning strategy, especially domain...

Human organs constantly undergo anatomical changes due to a complex mix of short-term (e.g., heartbeat) and long-term aging) factors. Evidently, prior knowledge these factors will be beneficial when modeling their future state, i.e., via image generation. However, most the medical generation tasks only rely on input from single image, thus ignoring sequential dependency even longitudinal data is available. Sequence-aware deep generative models, where model sequence ordered timestamped...