Enhancing schizophrenia phenotype prediction from genotype data through knowledge-driven deep neural network models.

This article explores deep learning model design, drawing inspiration from the omnigenic model and genetic heterogeneity concepts, to improve schizophrenia prediction using genotype data. It introduces an innovative three-step approach leveraging neural networks' capabilities to efficiently handle genetic interactions. A locally connected network initially routes input data from variants to their corresponding genes. The second step employs an Encoder-Decoder to capture relationships among identified genes. The final model integrates knowledge from the first two and incorporates a parallel component to consider the effects of additional genes. This expansion enhances prediction scores by considering a larger number of genes. Trained models achieved an average AUC of 0.83, surpassing other genotype-trained models and matching gene expression dataset-based approaches. Additionally, tests on held-out sets reported an average sensitivity of 0.72 and an accuracy of 0.76, aligning with schizophrenia heritability predictions. Moreover, the study addresses genetic heterogeneity challenges by considering diverse population subsets.

Convolutional neural network for brachial plexus segmentation at the interscalene level.

BACKGROUND: Regional anesthesia with ultrasound-guided brachial plexus block is widely used for patients undergoing shoulder and upper limb surgery, but needle misplacement can result in complications. The purpose of this study was to develop and validate a convolutional neural network (CNN) model for segmentation of the brachial plexus at the interscalene level. METHODS: This prospective study included patients who underwent ultrasound-guided brachial plexus block in the Anesthesiology Department of Beijing Jishuitan Hospital between October 2019 and June 2022. A Unet semantic segmentation model was developed to train the CNN to identify the brachial plexus features in the ultrasound images. The degree of overlap between the predicted segmentation and ground truth segmentation (manually drawn by experienced clinicians) was evaluated by calculation of the Dice index and Jaccard index. RESULTS: The final analysis included 502 images from 127 patients aged 41 ± 14 years-old (72 men, 56.7%). The mean Dice index was 0.748 ± 0.190, which was extremely close to the threshold level of 0.75 for good overlap between the predicted and ground truth segregations. The Jaccard index was 0.630 ± 0.213, which exceeded the threshold value of 0.5 for a good overlap. CONCLUSION: The CNN performed well at segregating the brachial plexus at the interscalene level. Further development could allow the CNN to be used to facilitate real-time identification of the brachial plexus during interscalene block administration. CLINICAL TRIAL REGISTRATION: The trial was registered prior to patient enrollment at the Chinese Clinical Trial Registry (ChiCTR2200055591), the site url is https://www.chictr.org.cn/ . The date of trial registration and patient enrollment is 14/01/2022.

Advancing maxillofacial prosthodontics by using pre-trained convolutional neural networks: Image-based classification of the maxilla.

PURPOSE: The study aimed to compare the performance of four pre-trained convolutional neural networks in recognizing seven distinct prosthodontic scenarios involving the maxilla, as a preliminary step in developing an artificial intelligence (AI)-powered prosthesis design system. MATERIALS AND METHODS: Seven distinct classes, including cleft palate, dentulous maxillectomy, edentulous maxillectomy, reconstructed maxillectomy, completely dentulous, partially edentulous, and completely edentulous, were considered for recognition. Utilizing transfer learning and fine-tuned hyperparameters, four AI models (VGG16, Inception-ResNet-V2, DenseNet-201, and Xception) were employed. The dataset, consisting of 3541 preprocessed intraoral occlusal images, was divided into training, validation, and test sets. Model performance metrics encompassed accuracy, precision, recall, F1 score, area under the receiver operating characteristic curve (AUC), and confusion matrix. RESULTS: VGG16, Inception-ResNet-V2, DenseNet-201, and Xception demonstrated comparable performance, with maximum test accuracies of 0.92, 0.90, 0.94, and 0.95, respectively. Xception and DenseNet-201 slightly outperformed the other models, particularly compared with InceptionResNet-V2. Precision, recall, and F1 scores exceeded 90% for most classes in Xception and DenseNet-201 and the average AUC values for all models ranged between 0.98 and 1.00. CONCLUSIONS: While DenseNet-201 and Xception demonstrated superior performance, all models consistently achieved diagnostic accuracy exceeding 90%, highlighting their potential in dental image analysis. This AI application could help work assignments based on difficulty levels and enable the development of an automated diagnosis system at patient admission. It also facilitates prosthesis designing by integrating necessary prosthesis morphology, oral function, and treatment difficulty. Furthermore, it tackles dataset size challenges in model optimization, providing valuable insights for future research.

A hybrid neural network based model for blood donation forecasting.

OBJECTIVE: To develop a hybrid neural network-based blood donation prediction method, via this predictive model, we can obtain the best estimate of whole blood in Beijing Tongzhou District Central Blood Station and help managers smoothly solve the allocation problem under fluctuating hospital demand and limited resources. METHOD: Inspired by the practical problems faced by blood stations providing transfusion services to several hospitals, a hybrid model based on a time-series prediction method and neural network, SARIMAX-TCN-LSTM is proposed for the prediction of daily whole blood donations. The experiment was performed at the central blood station in Tongzhou district, where we used whole blood donations from January 1, 2015, to November 14, 2021, as the subject, supplemented by meteorological and epidemic factors affecting blood donation, to predict daily blood donations for the next two weeks. RESULT: The hybrid model significantly outperformed the traditional time series forecasting method on multiple regression metrics, with twice as effective fitting as the baseline and a 33% reduction in Root Mean Squared Error (RMSE). Results indicate that the proposed model can improve the prediction accuracy of daily blood donations, and the co-validity of the structure was evidenced in an ablation experiment. CONCLUSION: Development and evaluation of a hybrid neural network-based model structure improve the prediction of daily blood donations. This intelligent forecasting method can help managers to overcome the challenges of sudden blood demand and contribute to the optimization of resource allocation tasks.

A 3D Radiomics-Based Artificial Neural Network Model for Benign Versus Malignant Vertebral Compression Fracture Classification in MRI.

To train an artificial neural network model using 3D radiomic features to differentiate benign from malignant vertebral compression fractures (VCFs) on MRI. This retrospective study analyzed sagittal T1-weighted lumbar spine MRIs from 91 patients (average age of 64.24 ± 11.75 years) diagnosed with benign or malignant VCFs from 2010 to 2019, of them 47 (51.6%) had benign VCFs and 44 (48.4%) had malignant VCFs. The lumbar fractures were three-dimensionally segmented and had their radiomic features extracted and selected with the wrapper method. The training set consisted of 100 fractured vertebral bodies from 61 patients (average age of 63.2 ± 12.5 years), and the test set was comprised of 30 fractured vertebral bodies from 30 patients (average age of 66.4 ± 9.9 years). Classification was performed with the multilayer perceptron neural network with a back-propagation algorithm. To validate the model, the tenfold cross-validation technique and an independent test set (holdout) were used. The performance of the model was evaluated using the average with a 95% confidence interval for the ROC AUC, accuracy, sensitivity, and specificity (considering the threshold = 0.5). In the internal validation test, the best model reached a ROC AUC of 0.98, an accuracy of 95% (95/100), a sensitivity of 93.5% (43/46), and specificity of 96.3% (52/54). In the validation with independent test set, the model achieved a ROC AUC of 0.97, an accuracy of 93.3% (28/30), a sensitivity of 93.3% (14/15), and a specificity of 93.3% (14/15). The model proposed in this study using radiomic features could differentiate benign from malignant vertebral compression fractures with excellent performance and is promising as an aid to radiologists in the characterization of VCFs.

Finding Distributed Needles in Neural Haystacks.

The human cortex encodes information in complex networks that can be anatomically dispersed and variable in their microstructure across individuals. Using simulations with neural network models, we show that contemporary statistical methods for functional brain imaging-including univariate contrast, searchlight multivariate pattern classification, and whole-brain decoding with L1 or L2 regularization-each have critical and complementary blind spots under these conditions. We then introduce the sparse-overlapping-sets (SOS) LASSO-a whole-brain multivariate approach that exploits structured sparsity to find network-distributed information-and show in simulation that it captures the advantages of other approaches while avoiding their limitations. When applied to fMRI data to find neural responses that discriminate visually presented faces from other visual stimuli, each method yields a different result, but existing approaches all support the canonical view that face perception engages localized areas in posterior occipital and temporal regions. In contrast, SOS LASSO uncovers a network spanning all four lobes of the brain. The result cannot reflect spurious selection of out-of-system areas because decoding accuracy remains exceedingly high even when canonical face and place systems are removed from the dataset. When used to discriminate visual scenes from other stimuli, the same approach reveals a localized signal consistent with other methods-illustrating that SOS LASSO can detect both widely distributed and localized representational structure. Thus, structured sparsity can provide an unbiased method for testing claims of functional localization. For faces and possibly other domains, such decoding may reveal representations more widely distributed than previously suspected.SIGNIFICANCE STATEMENT Brain systems represent information as patterns of activation over neural populations connected in networks that can be widely distributed anatomically, variable across individuals, and intermingled with other networks. We show that four widespread statistical approaches to functional brain imaging have critical blind spots in this scenario and use simulations with neural network models to illustrate why. We then introduce a new approach designed specifically to find radically distributed representations in neural networks. In simulation and in fMRI data collected in the well studied domain of face perception, the new approach discovers extensive signal missed by the other methods-suggesting that prior functional imaging work may have significantly underestimated the degree to which neurocognitive representations are distributed and variable across individuals.

Prediction of radiosensitivity and radiocurability using a novel supervised artificial neural network.

BACKGROUND: Radiotherapy has been widely used to treat various cancers, but its efficacy depends on the individual involved. Traditional gene-based machine-learning models have been widely used to predict radiosensitivity. However, there is still a lack of emerging powerful models, artificial neural networks (ANN), in the practice of gene-based radiosensitivity prediction. In addition, ANN may overfit and learn biologically irrelevant features. METHODS: We developed a novel ANN with Selective Connection based on Gene Patterns (namely ANN-SCGP) to predict radiosensitivity and radiocurability. We creatively used gene patterns (gene similarity or gene interaction information) to control the "on-off" of the first layer of weights, enabling the low-dimensional features to learn the gene pattern information. ANN-SCGP was trained and tested in 82 cell lines and 1,101 patients from the 11 pan-cancer cohorts. RESULTS: For survival fraction at 2 Gy, the root mean squared errors (RMSE) of prediction in ANN-SCGP was the smallest among all algorithms (mean RMSE: 0.1587-0.1654). For radiocurability, ANN-SCGP achieved the first and second largest C-index in the 12/20 and 4/20 tests, respectively. The low dimensional output of ANN-SCGP reproduced the patterns of gene similarity. Moreover, the pan-cancer analysis indicated that immune signals and DNA damage responses were associated with radiocurability. CONCLUSIONS: As a model including gene pattern information, ANN-SCGP had superior prediction abilities than traditional models. Our work provided novel insights into radiosensitivity and radiocurability.

Fuzzy Control of Pressure in a Water Supply Network Based on Neural Network System Modeling and IoT Measurements.

As hydroenergetic losses are inherent to water supply systems, they are a frequent issue which water utilities deal with every day. The control of network pressure is essential to reducing these losses, providing a quality supply to consumers, saving electricity and preserving piping from excess pressure. However, to obtain these benefits, it is necessary to overcome some difficulties such as sensing the pressure of geographically distant consumer units and developing a control logic that is capable of making use of the data from these sensors and, at the same time, a good solution in terms of cost benefit. Therefore, this work has the purpose of developing a pressure monitoring and control system for water supply networks, using the ESP8266 microcontroller to collect data from pressure sensors for the integrated ScadaLTS supervisory system via the REST API. The modeling of the plant was developed using artificial neural networks together with fuzzy pressure control, both designed using the Python language. The proposed method was tested by considering a pumping station and two reference units located in the city of João Pessoa, Brazil, in which there was an excess of pressure in the supply network and low performance from the old controls, during the night period from 12:00 a.m. to 6:00 a.m. The field results estimated 2.9% energy saving in relation to the previous form of control and a guarantee that the pressure in the network was at a healthy level.

An exploration of error-driven learning in simple two-layer networks from a discriminative learning perspective.

Error-driven learning algorithms, which iteratively adjust expectations based on prediction error, are the basis for a vast array of computational models in the brain and cognitive sciences that often differ widely in their precise form and application: they range from simple models in psychology and cybernetics to current complex deep learning models dominating discussions in machine learning and artificial intelligence. However, despite the ubiquity of this mechanism, detailed analyses of its basic workings uninfluenced by existing theories or specific research goals are rare in the literature. To address this, we present an exposition of error-driven learning - focusing on its simplest form for clarity - and relate this to the historical development of error-driven learning models in the cognitive sciences. Although historically error-driven models have been thought of as associative, such that learning is thought to combine preexisting elemental representations, our analysis will highlight the discriminative nature of learning in these models and the implications of this for the way how learning is conceptualized. We complement our theoretical introduction to error-driven learning with a practical guide to the application of simple error-driven learning models in which we discuss a number of example simulations, that are also presented in detail in an accompanying tutorial.

Test-retest repeatability of a deep learning architecture in detecting and segmenting clinically significant prostate cancer on apparent diffusion coefficient (ADC) maps.

OBJECTIVES: To evaluate short-term test-retest repeatability of a deep learning architecture (U-Net) in slice- and lesion-level detection and segmentation of clinically significant prostate cancer (csPCa: Gleason grade group > 1) using diffusion-weighted imaging fitted with monoexponential function, ADCm. METHODS: One hundred twelve patients with prostate cancer (PCa) underwent 2 prostate MRI examinations on the same day. PCa areas were annotated using whole mount prostatectomy sections. Two U-Net-based convolutional neural networks were trained on three different ADCm b value settings for (a) slice- and (b) lesion-level detection and (c) segmentation of csPCa. Short-term test-retest repeatability was estimated using intra-class correlation coefficient (ICC(3,1)), proportionate agreement, and dice similarity coefficient (DSC). A 3-fold cross-validation was performed on training set (N = 78 patients) and evaluated for performance and repeatability on testing data (N = 34 patients). RESULTS: For the three ADCm b value settings, repeatability of mean ADCm of csPCa lesions was ICC(3,1) = 0.86-0.98. Two CNNs with U-Net-based architecture demonstrated ICC(3,1) in the range of 0.80-0.83, agreement of 66-72%, and DSC of 0.68-0.72 for slice- and lesion-level detection and segmentation of csPCa. Bland-Altman plots suggest that there is no systematic bias in agreement between inter-scan ground truth segmentation repeatability and segmentation repeatability of the networks. CONCLUSIONS: For the three ADCm b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility. KEY POINTS: • For the three ADCm b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. • The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility.

Precise whole liver automatic segmentation and quantification of PDFF and R2* on MR images.

OBJECTIVE: To automate the segmentation of whole liver parenchyma on multi-echo chemical shift encoded (MECSE) MR examinations using convolutional neural networks (CNNs) to seamlessly quantify precise organ-related imaging biomarkers such as the fat fraction and iron load. METHODS: A retrospective multicenter collection of 183 MECSE liver MR examinations was conducted. An encoder-decoder CNN was trained (107 studies) following a 5-fold cross-validation strategy to improve the model performance and ensure lack of overfitting. Proton density fat fraction (PDFF) and R2* were quantified on both manual and CNN segmentation masks. Different metrics were used to evaluate the CNN performance over both unseen internal (46 studies) and external (29 studies) validation datasets to analyze reproducibility. RESULTS: The internal test showed excellent results for the automatic segmentation with a dice coefficient (DC) of 0.93 ± 0.03 and high correlation between the quantification done with the predicted mask and the manual segmentation (rPDFF = 1 and rR2* = 1; p values < 0.001). The external validation was also excellent with a different vendor but the same magnetic field strength, proving the generalization of the model to other manufacturers with DC of 0.94 ± 0.02. Results were lower for the 1.5-T MR same vendor scanner with DC of 0.87 ± 0.06. Both external validations showed high correlation in the quantification (rPDFF = 1 and rR2* = 1; p values < 0.001). In both internal and external validation datasets, the relative error for the PDFF and R2* quantification was below 4% and 1% respectively. CONCLUSION: Liver parenchyma can be accurately segmented with CNN in a vendor-neutral virtual approach, allowing to obtain reproducible automatic whole organ virtual biopsies. KEY POINTS: • Whole liver parenchyma can be automatically segmented using convolutional neural networks. • Deep learning allows the creation of automatic pipelines for the precise quantification of liver-related imaging biomarkers such as PDFF and R2*. • MR "virtual biopsy" can become a fast and automatic procedure for the assessment of chronic diffuse liver diseases in clinical practice.

Cholinergic Switch between Two Types of Slow Waves in Cerebral Cortex.

Sleep slow waves are known to participate in memory consolidation, yet slow waves occurring under anesthesia present no positive effects on memory. Here, we shed light onto this paradox, based on a combination of extracellular recordings in vivo, in vitro, and computational models. We find two types of slow waves, based on analyzing the temporal patterns of successive slow-wave events. The first type is consistently observed in natural slow-wave sleep, while the second is shown to be ubiquitous under anesthesia. Network models of spiking neurons predict that the two slow wave types emerge due to a different gain on inhibitory versus excitatory cells and that different levels of spike-frequency adaptation in excitatory cells can account for dynamical distinctions between the two types. This prediction was tested in vitro by varying adaptation strength using an agonist of acetylcholine receptors, which demonstrated a neuromodulatory switch between the two types of slow waves. Finally, we show that the first type of slow-wave dynamics is more sensitive to external stimuli, which can explain how slow waves in sleep and anesthesia differentially affect memory consolidation, as well as provide a link between slow-wave dynamics and memory diseases.

The four-minute approach revisited: accelerating MRI-based multi-factorial age estimation.

OBJECTIVES: This feasibility study aimed to investigate the reliability of multi-factorial age estimation based on MR data of the hand, wisdom teeth and the clavicles with reduced acquisition time. METHODS: The raw MR data of 34 volunteers-acquired on a 3T system and using acquisition times (TA) of 3:46 min (hand), 5:29 min (clavicles) and 10:46 min (teeth)-were retrospectively undersampled applying the commercially available CAIPIRINHA technique. Automatic and radiological age estimation methods were applied to the original image data as well as undersampled data to investigate the reliability of age estimates with decreasing acquisition time. Reliability was investigated determining standard deviation (SSD) and mean (MSD) of signed differences, intra-class correlation (ICC) and by performing Bland-Altman analysis. RESULTS: Automatic age estimation generally showed very high reliability (SSD < 0.90 years) even for very short acquisition times (SSD ≈ 0.20 years for a total TA of 4 min). Radiological age estimation provided highly reliable results for images of the hand (ICC ≥ 0.96) and the teeth (ICC ≥ 0.79) for short acquisition times (TA = 16 s for the hand, TA = 2:21 min for the teeth), imaging data of the clavicles allowed for moderate acceleration (TA = 1:25 min, ICC ≥ 0.71). CONCLUSIONS: The results demonstrate that reliable multi-factorial age estimation based on MRI of the hand, wisdom teeth and the clavicles can be performed using images acquired with a total acquisition time of 4 min.

Differentiation of Small (≤ 4 cm) Renal Masses on Multiphase Contrast-Enhanced CT by Deep Learning.

OBJECTIVE. This study evaluated the utility of a deep learning method for determining whether a small (≤ 4 cm) solid renal mass was benign or malignant on multiphase contrast-enhanced CT. MATERIALS AND METHODS. This retrospective study included 1807 image sets from 168 pathologically diagnosed small (≤ 4 cm) solid renal masses with four CT phases (unenhanced, corticomedullary, nephrogenic, and excretory) in 159 patients between 2012 and 2016. Masses were classified as malignant (n = 136) or benign (n = 32). The dataset was randomly divided into five subsets: four were used for augmentation and supervised training (48,832 images), and one was used for testing (281 images). The Inception-v3 architecture convolutional neural network (CNN) model was used. The AUC for malignancy and accuracy at optimal cutoff values of output data were evaluated in six different CNN models. Multivariate logistic regression analysis was also performed. RESULTS. Malignant and benign lesions showed no significant difference of size. The AUC value of corticomedullary phase was higher than that of other phases (corticomedullary vs excretory, p = 0.022). The highest accuracy (88%) was achieved in corticomedullary phase images. Multivariate analysis revealed that the CNN model of corticomedullary phase was a significant predictor for malignancy compared with other CNN models, age, sex, and lesion size. CONCLUSION. A deep learning method with a CNN allowed acceptable differentiation of small (≤ 4 cm) solid renal masses in dynamic CT images, especially in the corticomedullary image model.

The formation and use of hierarchical cognitive maps in the brain: A neural network model.

Many researchers have tried to model how environmental knowledge is learned by the brain and used in the form of cognitive maps. However, previous work was limited in various important ways: there was little consensus on how these cognitive maps were formed and represented, the planning mechanism was inherently limited to performing relatively simple tasks, and there was little consideration of how these mechanisms would scale up. This paper makes several significant advances. Firstly, the planning mechanism used by the majority of previous work propagates a decaying signal through the network to create a gradient that points towards the goal. However, this decaying signal limited the scale and complexity of tasks that can be solved in this manner. Here we propose several ways in which a network can can self-organize a novel planning mechanism that does not require decaying activity. We also extend this model with a hierarchical planning mechanism: a layer of cells that identify frequently-used sequences of actions and reuse them to significantly increase the efficiency of planning. We speculate that our results may explain the apparent ability of humans and animals to perform model-based planning on both small and large scales without a noticeable loss of efficiency.

Pervasive Lying Posture Tracking.

Automated lying-posture tracking is important in preventing bed-related disorders, such as pressure injuries, sleep apnea, and lower-back pain. Prior research studied in-bed lying posture tracking using sensors of different modalities (e.g., accelerometer and pressure sensors). However, there remain significant gaps in research regarding how to design efficient in-bed lying posture tracking systems. These gaps can be articulated through several research questions, as follows. First, can we design a single-sensor, pervasive, and inexpensive system that can accurately detect lying postures? Second, what computational models are most effective in the accurate detection of lying postures? Finally, what physical configuration of the sensor system is most effective for lying posture tracking? To answer these important research questions, in this article we propose a comprehensive approach for designing a sensor system that uses a single accelerometer along with machine learning algorithms for in-bed lying posture classification. We design two categories of machine learning algorithms based on deep learning and traditional classification with handcrafted features to detect lying postures. We also investigate what wearing sites are the most effective in the accurate detection of lying postures. We extensively evaluate the performance of the proposed algorithms on nine different body locations and four human lying postures using two datasets. Our results show that a system with a single accelerometer can be used with either deep learning or traditional classifiers to accurately detect lying postures. The best models in our approach achieve an F1 score that ranges from 95.2% to 97.8% with a coefficient of variation from 0.03 to 0.05. The results also identify the thighs and chest as the most salient body sites for lying posture tracking. Our findings in this article suggest that, because accelerometers are ubiquitous and inexpensive sensors, they can be a viable source of information for pervasive monitoring of in-bed postures.

Prediction of local relapse and distant metastasis in patients with definitive chemoradiotherapy-treated cervical cancer by deep learning from [18F]-fluorodeoxyglucose positron emission tomography/computed tomography.

BACKGROUND: We designed a deep learning model for assessing 18F-FDG PET/CT for early prediction of local and distant failures for patients with locally advanced cervical cancer. METHODS: All 142 patients with cervical cancer underwent 18F-FDG PET/CT for pretreatment staging and received allocated treatment. To augment the amount of image data, each tumor was represented as 11 slice sets each of which contains 3 2D orthogonal slices to acquire a total of 1562 slice sets. In each round of k-fold cross-validation, a well-trained proposed model and a slice-based optimal threshold were derived from a training set and used to classify each slice set in the test set into the categories of with or without local or distant failure. The classification results of each tumor were aggregated to summarize a tumor-based prediction result. RESULTS: In total, 21 and 26 patients experienced local and distant failures, respectively. Regarding local recurrence, the tumor-based prediction result summarized from all test sets demonstrated that the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 71%, 93%, 63%, 95%, and 89%, respectively. The corresponding values for distant metastasis were 77%, 90%, 63%, 95%, and 87%, respectively. CONCLUSION: This is the first study to use deep learning model for assessing 18F-FDG PET/CT images which is capable of predicting treatment outcomes in cervical cancer patients. KEY POINTS: • This is the first study to use deep learning model for assessing 18 F-FDG PET/CT images which is capable of predicting treatment outcomes in cervical cancer patients. • All 142 patients with cervical cancer underwent 18 F-FDG PET/CT for pretreatment staging and received allocated treatment. To augment the amount of image data, each tumor was represented as 11 slice sets each of which contains 3 2D orthogonal slices to acquire a total of 1562 slice sets. • For local recurrence, all test sets demonstrated that the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 71%, 93%, 63%, 95%, and 89%, respectively. The corresponding values for distant metastasis were 77%, 90%, 63%, 95%, and 87%, respectively.

An approach for medical event detection in Chinese clinical notes of electronic health records.

BACKGROUND: Medical event detection in narrative clinical notes of electronic health records (EHRs) is a task designed for reading text and extracting information. Most of the previous work of medical event detection treats the task as extracting concepts at word granularity, which omits the overall structural information of the clinical notes. In this work, we treat each clinical note as a sequence of short sentences and propose an end-to-end deep neural network framework. METHODS: We redefined the task as a sequence labelling task at short sentence granularity, and proposed a novel tag system correspondingly. The dataset were derived from a third-level grade-A hospital, consisting of 2000 annotated clinical notes according to our proposed tag system. The proposed end-to-end deep neural network framework consists of a feature extractor and a sequence labeller, and we explored different implementations respectively. We additionally proposed a smoothed Viterbi decoder as sequence labeller without additional parameter training, which can be a good alternative to conditional random field (CRF) when computing resources are limited. RESULTS: Our sequence labelling models were compared to four baselines which treat the task as text classification of short sentences. Experimental results showed that our approach significantly outperforms the baselines. The best result was obtained by using the convolutional neural networks (CNNs) feature extractor and the sequential CRF sequence labeller, achieving an accuracy of 92.6%. Our proposed smoothed Viterbi decoder achieved a comparable accuracy of 90.07% with reduced training parameters, and brought more balanced performance across all categories, which means better generalization ability. CONCLUSIONS: Evaluated on our annotated dataset, the comparison results demonstrated the effectiveness of our approach for medical event detection in Chinese clinical notes of EHRs. The best feature extractor is the CNNs feature extractor, and the best sequence labeller is the sequential CRF decoder. And it was empirically verified that our proposed smoothed Viterbi decoder could bring better generalization ability while achieving comparable performance to the sequential CRF decoder.

Objective auditory brainstem response classification using machine learning.

OBJECTIVE: The objective of this study was to use machine learning in the form of a deep neural network to objectively classify paired auditory brainstem response waveforms into either: 'clear response', 'inconclusive' or 'response absent'. DESIGN: A deep convolutional neural network was constructed and fine-tuned using stratified 10-fold cross-validation on 190 paired ABR waveforms. The final model was evaluated on a test set of 42 paired waveforms. STUDY SAMPLE: The full dataset comprised 232 paired ABR waveforms recorded from eight normal-hearing individuals. The dataset was obtained from the PhysioBank database. The paired waveforms were independently labelled by two audiological scientists in order to train the network and evaluate its performance. RESULTS: The trained neural network was able to classify paired ABR waveforms with 92.9% accuracy. The sensitivity and the specificity were 92.9% and 96.4%, respectively. CONCLUSIONS: This neural network may have clinical utility in assisting clinicians with waveform classification for the purpose of hearing threshold estimation. Further evaluation using a large clinically obtained dataset would provide further validation with regard to the clinical potential of the neural network in diagnostic adult testing, newborn testing and in automated newborn hearing screening.

Real and predicted mortality under health spending constraints in Italy: a time trend analysis through artificial neural networks.

BACKGROUND: After 2008 global economic crisis, Italian governments progressively reduced public healthcare financing. Describing the time trend of health outcomes and health expenditure may be helpful for policy makers during the resources' allocation decision making process. The aim of this paper is to analyze the trend of mortality and health spending in Italy and to investigate their correlation in consideration of the funding constraints experienced by the Italian national health system (SSN). METHODS: We conducted a 20-year time-series study. Secondary data has been extracted from a national, institution based and publicly accessible retrospective database periodically released by the Italian Institute of Statistics. Age standardized all-cause mortality rate (MR) and health spending (Directly Provided Services - DPS, Agreed-Upon Services - TAUS, and private expenditure) were reviewed. Time trend analysis (1995-2014) through OLS and Multilayer Feed-forward Neural Networks (MFNN) models to forecast mortality and spending trend was performed. The association between healthcare expenditure and MR was analyzed through a fixed effect regression model. We then repeated MFNN time trend forecasting analyses on mortality by adding the spending item resulted significantly related with MR in the fixed effect analyses. RESULTS: DPS and TAUS decreased since 2011. There was a mismatch in mortality rates between real and predicted values. DPS resulted significantly associated to mortality (p < 0.05). In repeated mortality forecasting analysis, predicted MR was found to be lower when considering the pre-constraints health spending trend. CONCLUSIONS: Between 2011 and 2014, Italian public health spending items showed a reduction when compared to prior years. Spending on services directly provided free of charge appears to be the financial driving force of the Italian public health system. The overall mortality was found to be higher than the predicted trend and this scenario may be partially attributable to the healthcare funding constraints experienced by the SSN.