Artificial Intelligence Learning Semantics via External Resources for Classifying Diagnosis Codes in Discharge Notes.

We aimed to assess the ability of deep learning (DL) and support vector machine (SVM) to detect a nonperfusion area (NPA) caused by retinal vein occlusion (RVO) with optical coherence tomography angiography (OCTA) images. The study included 322 OCTA images (normal: 148; NPA owing to RVO: 174 [128 branch RVO images and 46 central RVO images]). Training to construct the DL model using deep convolutional neural network (DNN) algorithms was provided using OCTA images. The SVM used a scikit-learn library with a radial basis function kernel. The area under the curve (AUC), sensitivity and specificity for detecting an NPA were examined. We compared the diagnostic ability (sensitivity, specificity and average required time) between the DNN, SVM and seven ophthalmologists. Heat maps were generated. With regard to the DNN, the mean AUC, sensitivity, specificity and average required time for distinguishing RVO OCTA images with an NPA from normal OCTA images were 0.986, 93.7%, 97.3% and 176.9 s, respectively. With regard to SVM, the mean AUC, sensitivity, and specificity were 0.880, 79.3%, and 81.1%, respectively. With regard to the seven ophthalmologists, the mean AUC, sensitivity, specificity and average required time were 0.962, 90.8%, 89.2%, and 700.6 s, respectively. The DNN focused on the foveal avascular zone and NPA in heat maps. The performance of the DNN was significantly better than that of SVM in all parameters (p < 0.01, all) and that of the ophthalmologists in AUC and specificity (p < 0.01, all). The combination of DL and OCTA images had high accuracy for the detection of an NPA, and it might be useful in clinical practice and retinal screening.

The subfornical organ (SFO), a sensory circumventricular organ lacking the normal blood-brain barrier with well documented roles in cardiovascular regulation, has recently been identified as a potential site at which the adipokine, leptin, may act to influence central autonomic pathways. Systemic and central leptin administration has been shown to increase blood pressure and it has been suggested that selective leptin resistance contributes to obesity-related hypertension. Given the relationship between obesity and hypertension, the present study aimed to investigate the cardiovascular consequences of the direct administration of leptin into the SFO of young lean rats and in the diet-induced obesity (DIO) rat model, which has been shown to be leptin-resistant. Leptin administration (500 fmol) directly into the SFO of young rats resulted in rapid decreases in blood pressure (BP) [mean area under the curve (AUC) = -677.8 ± 167.1 mmHg*s; n = 9], without an effect on heart rate (mean AUC = -21.2 ± 13.4 beats; n = 9), and these effects were found to be dose-related as microinjection of 5 pmol of leptin into the SFO had a larger effect on BP (mean AUC = -972.3 ± 280.1 mmHg*s; n = 4). These BP effects were also shown to be site-specific as microinjection of leptin into non-SFO regions or into the ventricle was without effect on BP (non-SFO: mean AUC = -22.4 ± 55.3 mmHg*s; n = 4; ventricle: mean AUC = 194.0 ± 173.0 mmHg*s; n = 6). By contrast, microinjection of leptin into leptin-resistant DIO rats was without effect on BP (mean AUC = 205.2 ± 75.1 mmHg*s; n = 4). These observations suggest that the SFO may be an important relay centre through which leptin, in normal weight, leptin responsive rats, acts to maintain BP within normal physiological limits through descending autonomic pathways involved in cardiovascular control and that, in obese, leptin-resistant, rats leptin no longer influences SFO neurones, resulting in an elevated BP, thus contributing to obesity-related hypertension.

Contrast-enhanced mammography (CEM) is a promising breast imaging technique. A limited number of studies have focused on the radiomics analysis of CEM. We intended to explore whether a model constructed with both clinical and radiomics features of CEM can better classify benign and malignant breast lesions. This retrospective, double-center study included women who underwent CEM between August 2017 and February 2020. The data from Center 1 were used as training set and the data from Center 2 were used as external testing set (training: testing =2:1). Models were constructed with the clinical, radiomics, and clinical + radiomics features of CEM. The clinical features included patient age and clinical image features interpreted by the radiologists. The radiomics features were extracted from high-energy (HE), low-energy (LE), and dual-energy subtraction (DES) images of CEM. The Mann-Whitney U test, Pearson correlation and Boruta's approach were used to select the radiomics features. Random Forest (RF) and logistic regression were used to establish the models. For the testing set, the areas under the curve (AUCs) and 95% confidence intervals (CIs) were employed to evaluate the performance of the models. For the training set, the mean AUCs were obtained by performing internal validation for 100 iterations and then compared by the Kruskal-Wallis and Mann-Whitney U tests. A total of 226 women (mean age: 47.4±10.1 years) with 226 pathologically proven breast lesions (101 benign; 125 malignant) were included. For the external testing set, the AUCs were 0.964 (95% CI: 0.918-1.000) for the combined model, 0.947 (95% CI: 0.891-0.997) for the radiomics model, and 0.882 (95% CI: 0.803-0.962) for the clinical model. In the internal validation process, the combined model achieved a mean AUC of 0.934±0.030, which was significantly higher than those of the radiomics (mean AUC =0.921±0.031, adjusted P<0.050) and clinical models (mean AUC =0.907±0.036; adjusted P<0.050). Incorporating both clinical and radiomics features of CEM may achieve better classification results for breast lesions.

Freezing of gait (FOG) is one of the most debilitating motor symptoms experienced by patients with Parkinson's disease (PD). FOG detection is possible using acceleration data from wearable sensors, and a convolutional neural network (CNN) is often used to determine the presence of FOG epochs. We compared the performance of a standard CNN for the detection of FOG with two more complex networks, which are well suited for time series data, the MiniRocket and the InceptionTime. We combined acceleration data of people with PD across four studies. The final data set was split into a training (80%) and hold-out test (20%) set. A fifth study was included as an unseen test set. The data were windowed (2 s) and five-fold cross-validation was applied. The CNN, MiniRocket, and InceptionTime models were evaluated using a receiver operating characteristic (ROC) curve and its area under the curve (AUC). Multiple sensor configurations were evaluated for the best model. The geometric mean was subsequently calculated to select the optimal threshold. The selected model and threshold were evaluated on the hold-out and unseen test set. A total of 70 participants (23.7 h, 9% FOG) were included in this study for training and testing, and in addition, 10 participants provided an unseen test set (2.4 h, 11% FOG). The CNN performed best (AUC = 0.86) in comparison to the InceptionTime (AUC = 0.82) and MiniRocket (AUC = 0.76) models. For the CNN, we found a similar performance for a seven-sensor configuration (lumbar, upper and lower legs and feet; AUC = 0.86), six-sensor configuration (upper and lower legs and feet; AUC = 0.87), and two-sensor configuration (lower legs; AUC = 0.86). The optimal threshold of 0.45 resulted in a sensitivity of 77% and a specificity of 58% for the hold-out set (AUC = 0.72), and a sensitivity of 85% and a specificity of 68% for the unseen test set (AUC = 0.90). We confirmed that deep learning can be used to detect FOG in a large, heterogeneous dataset. The CNN model outperformed more complex networks. This model could be employed in future personalized interventions, with the ultimate goal of using automated FOG detection to trigger real-time cues to alleviate FOG in daily life.

Aims: This study aimed to explore the effect of training set diversity on the performance of deep learning models for predicting early gastric cancer (EGC) under endoscopy. Methods: Images of EGC and non-cancerous lesions under narrow-band imaging (ME-NBI) and magnifying blue laser imaging (ME-BLI) were retrospectively collected. Training set 1 was composed of 150 non-cancerous and 309 EGC ME-NBI images, training set 2 was composed of 1505 non-cancerous and 309 EGC ME-BLI images, and training set 3 was the combination of training set 1 and 2. Test set 1 was composed of 376 non-cancerous and 1052 EGC ME-NBI images, test set 2 consisted of 529 non-cancerous and 71 EGC ME-BLI images, and test set 3 was the combination of test set 1 and test set 2. Three deep learning models, convolutional neural network (CNN) 1, CNN 2 and CNN 3 (CNN 1, CNN 2 and CNN 3 were independently trained using training set 1, training set 2 and training set 3, respectively), were constructed, and their performances on each test set were respectively evaluated. One hundred and thirty-eight ME-NBI videos and 17 ME-BLI videos were further collected to evaluate and compare the performance of each model in real time. Results: On the whole, the performance of CNN 3 was the best. The accuracy (Acc), sensitivity (Sn), specificity (Sp) and area under the curve (AUC) of test set 1 in CNN 3 were 87.89% (1255/1428), 90.96% (342/376), 86.79% (913/1052) and 94.60%, respectively. The Acc, Sn, Sp and AUC of test set 2 in CNN 3 were 95% (570/600), 97.92% (518/529), 73.24% (52/71) and 90.93% respectively. The Acc, Sn, Sp and AUC of test set 3 in CNN 3 were 89.99% (1825/2028), 95.03% (860/905), 85.93% (965/1123) and 94.89%, respectively. The performance of CNN 3 was also the best in videos test set. The Acc, Sn and Sp of videos test set in CNN 3 were 91.03% (142/156), 90.58% (125/138) and 94.44% (17/18), respectively. Conclusions: The deep learning model with the most diverse training data has the best diagnostic effect.

Background Deep learning may help to improve computer-aided detection of volume (CADv) measurement of pulmonary nodules at chest CT. Purpose To determine the efficacy of a deep learning method for improving CADv for measuring the solid and ground-glass opacity (GGO) volumes of a nodule, doubling time (DT), and the change in volume at chest CT. Materials and Methods From January 2014 to December 2016, patients with pulmonary nodules at CT were retrospectively reviewed. CADv without and with a convolutional neural network (CNN) automatically determined total nodule volume change per day and DT. Area under the curves (AUCs) on a per-nodule basis and diagnostic accuracy on a per-patient basis were compared among all indexes from CADv with and without CNN for differentiating benign from malignant nodules. Results The CNN training set was 294 nodules in 217 patients, the validation set was 41 nodules in 32 validation patients, and the test set was 290 nodules in 188 patients. A total of 170 patients had 290 nodules (mean size ± standard deviation, 11 mm ± 5; range, 4-29 mm) diagnosed as 132 malignant nodules and 158 benign nodules. There were 132 solid nodules (46%), 106 part-solid nodules (36%), and 52 ground-glass nodules (18%). The test set results showed that the diagnostic performance of the CNN with CADv for total nodule volume change per day was larger than DT of CADv with CNN (AUC, 0.94 [95% confidence interval {CI}: 0.90, 0.96] vs 0.67 [95% CI: 0.60, 0.74]; P < .001) and CADv without CNN (total nodule volume change per day: AUC, 0.69 [95% CI: 0.62, 0.75]; P < .001; DT: AUC, 0.58 [95% CI: 0.51, 0.65]; P < .001). The accuracy of total nodule volume change per day of CADv with CNN was significantly higher than that of CADv without CNN (P < .001) and DT of both methods (P < .001). Conclusion Convolutional neural network is useful for improving accuracy of computer-aided detection of volume measurement and nodule differentiation capability at CT for patients with pulmonary nodules. © RSNA, 2020 Online supplemental material is available for this article.

Background and Aims Large language models (LLMs) have gained significant attention in the field of natural language processing (NLP), marking a shift from traditional techniques like Term Frequency-Inverse Document Frequency (TF-IDF). We developed a traditional NLP model to predict arteriovenous fistula (AVF) failure within next 30 days using clinical notes. The goal of this analysis was to investigate whether LLMs would outperform traditional NLP techniques, specifically in the context of predicting AVF failure within the next 30 days using clinical notes. Method We defined AVF failure as the change in status from active to permanently unusable status or temporarily unusable status. We used data from a large kidney care network from January 2021 to December 2021. Two models were created using LLMs and traditional TF-IDF technique. We used “distilbert-base-uncased”, a distilled version of BERT base model [1], and compared its performance with traditional TF-IDF-based NLP techniques. The dataset was randomly divided into 60% training, 20% validation and 20% test dataset. The test data, comprising of unseen patients’ data was used to evaluate the performance of the model. Both models were evaluated using metrics such as area under the receiver operating curve (AUROC), accuracy, sensitivity, and specificity. Results The incidence of 30 days AVF failure rate was 2.3% in the population. Both LLMs and traditional showed similar overall performance as summarized in Table 1. Notably, LLMs showed marginally better performance in certain evaluation metrics. Both models had same AUROC of 0.64 on test data. The accuracy and balanced accuracy for LLMs were 72.9% and 59.7%, respectively, compared to 70.9% and 59.6% for the traditional TF-IDF approach. In terms of specificity, LLMs scored 73.2%, slightly higher than the 71.2% observed for traditional NLP methods. However, LLMs had a lower sensitivity of 46.1% compared to 48% for traditional NLP. However, it is worth noting that training on LLMs took considerably longer than TF-IDF. Moreover, it also used higher computational resources such as utilization of graphics processing units (GPU) instances in cloud-based services, leading to higher cost. Conclusion In our study, we discovered that advanced LLMs perform comparably to traditional TF-IDF modeling techniques in predicting the failure of AVF. Both models demonstrated identical AUROC. While specificity was higher in LLMs compared to traditional NLP, sensitivity was higher in traditional NLP compared to LLMs. LLM was fine-tuned with a limited dataset, which could have influenced its performance to be similar to that of traditional NLP methods. This finding suggests that while LLMs may excel in certain scenarios, such as performing in-depth sentiment analysis of patient data for complex tasks, their effectiveness is highly dependent on the specific use case. It is crucial to weigh the benefits against the resources required for LLMs, as they can be significantly more resource-intensive and costly compared to traditional TF-IDF methods. This highlights the importance of a use-case-driven approach in selecting the appropriate NLP technique for healthcare applications.

ABSTRACTPurpose/Aim: To evaluate the impact of ocular parameters on the visual performance achieved with the multifocal intraocular lens (IOL) Bi-Flex M 677MY.Materials and Methods: About 26 subjects were included in the current study. Several physiological variables were retrieved from the 3-month follow-up visit, including pupil diameter and distance from pupil center to the vertex normal of the anterior cornea (µ). These variables were also obtained in the preoperative visit. Binocular and monocular visual acuity defocus curves were measured at 1 and 3 months after surgery, respectively. The monocular Area Under the Curve (AUC) was computed along the total range (Total Area Under the Curve (TAUC), +1.00 to −4.00 D) and for the ranges of Far (Far Area Under the Curve (FAUC), +0.50 to −0.50 D), (Intermediate Area Under the Curve (IAUC), −1.00 to −1.50 D) and Near vision (Near Area Under the Curve (NAUC), −2.00 to −4.00 D). Correlations between these areas and the postoperative physiological variables were assessed.Results: The mean µ was reduced from 0.21 to 0.10 mm after surgery, as well as pupil diameters, either photopic (−7.4%) or mesopic (−8.1%) (p < 0.05). The mean AUCs were 2.08 ± 0.74 for TAUC, 0.57 ± 0.17 for FAUC, 0.16 ± 0.09 for IAUC, and 0.81 ± 0.29 for NAUC. Significant correlations were found between NAUC and corneal power (r = −0.39, p = 0.05) as well as between IAUC and temporal decentration of the lens from vertex normal (ρ = −0.41, p = 0.04).Conclusions: The visual performance at near distance with the IOL evaluated improved in eyes with less corneal power. On the other hand, a slight temporal IOL decentration from vertex normal also improved intermediate visual acuity. The binocular defocus curve was similar to other trifocal IOLs.

We aim to evaluate the performance of radiomic biopsy (RB), best-fit bounding box (BB), and a deep-learning-based segmentation method called no-new-U-Net (nnU-Net), compared to the standard full manual (FM) segmentation method for predicting benign and malignant lung nodules using a computed tomography (CT) radiomic machine learning model. A total of 188 CT scans of lung nodules from 2 institutions were used for our study. One radiologist identified and delineated all 188 lung nodules, whereas a second radiologist segmented a subset () of these nodules. Both radiologists employed FM and RB segmentation methods. BB segmentations were generated computationally from the FM segmentations. The nnU-Net, a deep-learning-based segmentation method, performed automatic nodule detection and segmentation. The time radiologists took to perform segmentations was recorded. Radiomic features were extracted from each segmentation method, and models to predict benign and malignant lung nodules were developed. The Kruskal-Wallis and DeLong tests were used to compare segmentation times and areas under the curve (AUC), respectively. For the delineation of the FM, RB, and BB segmentations, the two radiologists required a median time (IQR) of 113 (54 to 251.5), 21 (9.25 to 38), and 16 (12 to 64.25) s, respectively (). In dataset 1, the mean AUC (95% CI) of the FM, RB, BB, and nnU-Net model were 0.964 (0.96 to 0.968), 0.985 (0.983 to 0.987), 0.961 (0.956 to 0.965), and 0.878 (0.869 to 0.888). In dataset 2, the mean AUC (95% CI) of the FM, RB, BB, and nnU-Net model were 0.717 (0.705 to 0.729), 0.919 (0.913 to 0.924), 0.699 (0.687 to 0.711), and 0.644 (0.632 to 0.657). Radiomic biopsy-based models outperformed FM and BB models in prediction of benign and malignant lung nodules in two independent datasets while deep-learning segmentation-based models performed similarly to FM and BB. RB could be a more efficient segmentation method, but further validation is needed.

Radiomics-based histological grading of pancreatic ductal adenocarcinoma using 18F-FDG PET/CT: A two-center study.

Virtual clinical trials (VCTs) require computer simulations of representative patients and images to evaluate and compare changes in performance of imaging technologies. The simulated images are usually interpreted by model observers whose performance depends upon the selection of imaging cases used in training evaluation models. This work proposes an efficient method to simulate and calibrate soft tissue lesions, which matches the detectability threshold of virtual and human readings. Anthropomorphic breast phantoms were used to evaluate the simulation of four mass models (I-IV) that vary in shape and composition of soft tissue. Ellipsoidal (I) and spiculated (II-IV) masses were simulated using composite voxels with partial volumes. Digital breast tomosynthesis projections and reconstructions of a clinical system were simulated. Channelized Hotelling observers (CHOs) were evaluated using reconstructed slices of masses that varied in shape, composition, and density of surrounded tissue. The detectability threshold of each mass model was evaluated using receiver operating characteristic (ROC) curves calculated with the CHO's scores. The area under the curve (AUC) of each calibrated mass model were within the 95% confidence interval (mean AUC [95% CI]) reported in a previous reader study (0.93 [0.89, 0.97]). The mean AUC [95% CI] obtained were 0.94 [0.93, 0.96], 0.92 [0.90, 0.93], 0.92 [0.90, 0.94], 0.93 [0.92, 0.95] for models I to IV, respectively. The mean AUC results varied substantially as a function of shape, composition, and density of surrounded tissue. For successful VCTs, lesions composed of soft tissue should be calibrated to simulate imaging cases that match the case difficulty predicted by human readers. Lesion composition, shape, and size are parameters that should be carefully selected to calibrate VCTs.

Nowadays word embeddings, also known as word vectors, play an important role for many natural language processing (NLP) tasks. In general, these word embeddings are learned from unsupervised learning models (e.g. Word2Vec, GloVe) with a large unannotated corpus and they are independent with the task of their application. In this paper we aim to enrich word embeddings by adding more information from a specific task that is the aspect based sentiment analysis. We propose a model using a convolutional neural network that takes a labeled data set, the learned word embeddings from an unsupervised learning model (e.g. Word2Vec) as input and fine-tunes word embeddings to capture aspect category and sentiment information. We conduct experiments on restaurant review data (http://spidr-ursa.rutgers.edu/datasets/). Experimental results show that fine-tuned word embeddings outperform unsupervisedly learned word embeddings.

Background: Heart failure poses a significant global health challenge, with high rates of readmission and mortality. Accurate models to predict these outcomes are essential for effective patient management. This study investigates the impact of data pre-processing techniques on XGBoost model performance in predicting all-cause readmission and mortality among heart failure patients. Methods: A dataset of 168 features from 2008 heart failure patients was used. Pre-processing included handling missing values, categorical encoding, and standardization. Four imputation techniques were compared: Mean, Multivariate Imputation by Chained Equations (MICEs), k-nearest Neighbors (kNNs), and Random Forest (RF). XGBoost models were evaluated using accuracy, recall, F1-score, and Area Under the Curve (AUC). Robustness was assessed through 10-fold cross-validation. Results: The XGBoost model with kNN imputation, one-hot encoding, and standardization outperformed others, with an accuracy of 0.614, recall of 0.551, and F1-score of 0.476. The MICE-based model achieved the highest AUC (0.647) and mean AUC (0.65 ± 0.04) in cross-validation. All pre-processed models outperformed the default XGBoost model (AUC: 0.60). Conclusions: Data pre-processing, especially MICE with one-hot encoding and standardization, improves XGBoost performance in heart failure prediction. However, moderate AUC scores suggest further steps are needed to enhance predictive accuracy.

Plasma levels of 6-mercaptopurine were determined in 22 consecutive children with acute lymphoblastic leukemia on oral remission maintenance therapy during the time period of August 1984 to January 1988. Each child received the drug once daily for up to 3 years and was studied repeatedly (1-12 times). An HPLC method was used for drug analysis. We found large interpatient variations in the mean peak plasma concentration (range of 50-424 ng/ml) and in the mean area under the concentration vs. time from 0-4 h curve (range of 82-637 ng ml-1 h). There were also pronounced variations between different sampling occasions in the same patient. Nine of the 22 patients had complications during the maintenance therapy. Five children with a mean peak plasma level below 135 ng/ml and a mean area under the curve (AUC) value below 251 ng ml-1 h relapsed (three in the central nervous system and two in the bone marrow). Both children with a bone marrow relapse died. Relapse risk was related to the AUC (p less than 0.05). Four children with a mean peak plasma level above 166 ng/ml and a mean AUC value above 363 ng/ml/h developed severe myelotoxicity, which necessitated a temporary cessation of the maintenance therapy. In addition, two patient relapsed 6 and 11 months after termination of maintenance therapy. Their mean peak and AUC values were not low but the concentrations decreased markedly towards the end of the maintenance period. The results indicate that the plasma levels of 6-mercaptopurine, when determined repeatedly, might be of significance for the outcome of the remission maintenance treatment.

Convolutional neural networks (CNN) are multi-layer neural networks that are used to learn hierarchical data properties. In recent times, CNN has achieved remarkable advances in the architecture and computation of Natural Language Processing (NLP). The Word2vec technique is considered to introduce Word embeddings, which are used to improve the performance of a variety of Natural Language Processing (NLP) applications. It is a well-known technique for learning word embeddings, which are dense representations of words in a lower-dimensional vector space. Two prominent approaches are used for learning word embeddings, which are dense representations of words in a lower-dimensional vector space, are Continuous Bag-of-Words (CBOW) and Skip-Gram.