Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

A convolutional neural network architecture to enhance oximetry ability to diagnose pediatric obstructive sleep apnea.

In IEEE journal of biomedical and health informatics

This study aims at assessing the usefulness of deep learning to enhance the diagnostic ability of oximetry in the context of automated detection of pediatric obstructive sleep apnea (OSA). A total of 3196 blood oxygen saturation (SpO2) signals from children were used for this purpose. A convolutional neural network (CNN) architecture was trained using 20-min SpO2 segments from the training set (859 subjects) to estimate the number of apneic events. CNN hyperparameters were tuned using Bayesian optimization in the validation set (1402 subjects). This model was applied to three test sets composed of 312, 392, and 231 subjects from three independent databases, in which the apnea-hypopnea index (AHI) estimated for each subject (AHICNN) was obtained by aggregating the output of the CNN for each 20-min SpO2 segment. AHICNN outperformed the 3% oxygen desaturation index (ODI3), a clinical approach, as well as the AHI estimated by a conventional feature-engineering approach based on multi-layer perceptron (AHIMLP). Specifically, AHICNN reached higher four-class Cohen's kappa in the three test databases than ODI3 (0.515 vs 0.417, 0.422 vs 0.372, and 0.423 vs 0.369) and AHIMLP (0.515 vs 0.377, 0.422 vs 0.381, and 0.423 vs 0.306). In addition, our proposal outperformed state-of-the-art studies, particularly for the AHI severity cutoffs of 5 e/h and 10 e/h. This suggests that the information automatically learned from the SpO2 signal by deep-learning techniques helps to enhance the diagnostic ability of oximetry in the context of pediatric OSA.

Vaquerizo-Villar Fernando, Alvarez Daniel, Kheirandish-Gozal Leila, Gutierrez-Tobal Gonzalo Cesar, Barroso-Garcia Veronica, Santamaria-Vazquez Eduardo, Del Campo Felix, Gozal David, Hornero Roberto


General General

A new estimation method for the biological interaction predicting problems.

In IEEE/ACM transactions on computational biology and bioinformatics

For the past decades, computational methods have been developed to predict various interactions in biological problems. Usually, these methods treated the predicting problems as semi-supervised problems or positive-unlabeled(PU) learning problems. Researchers focused on the prediction of unlabeled samples and hoped to find novel interactions in the datasets they collected. However, most of the computational methods could only predict a small proportion of undiscovered interactions and the total number was unknown. In this paper, we developed an estimation method with deep learning to calculate the number of undiscovered interactions in the unlabeled samples, derived its asymptotic interval estimation, and applied it to the compound synergism dataset, drug-target interaction(DTI) dataset, and MicroRNA-disease interaction dataset successfully. Moreover, this method could reveal which dataset contained more undiscovered interactions and would be a guidance for the experimental validation. Furthermore, we compared our method with some mixture proportion estimators and demonstrated the efficacy of our method. Finally, we proved that AUC and AUPR were related to the number of undiscovered interactions, which was regarded as another evaluation indicator for the computational methods.

Zhou Lewei, Tang Yucong, Yan Guiying


General General

Skin complications of diabetes mellitus revealed by polarized hyperspectral imaging and machine learning.

In IEEE transactions on medical imaging ; h5-index 74.0

Aging and diabetes lead to protein glycation and cause dysfunction of collagen-containing tissues. The accompanying structural and functional changes of collagen significantly contribute to the development of various pathological malformations affecting the skin, blood vessels, and nerves, causing a number of complications, increasing disability risks and threat to life. In fact, no methods of non-invasive assessment of glycation and associated metabolic processes in biotissues or prediction of possible skin complications, e.g., ulcers, currently exist for endocrinologists and clinical diagnosis. In this publication, utilizing emerging photonics-based technology, innovative solutions in machine learning, and definitive physiological characteristics, we introduce a diagnostic approach capable of evaluating the skin complications of diabetes mellitus at the very earlier stage. The results of the feasibility studies, as well as the actual tests on patients with diabetes and healthy volunteers, clearly show the ability of the approach to differentiate diabetic and control groups. Furthermore, the developed in-house polarization-based hyperspectral imaging technique accomplished with the implementation of the artificial neural network provides new horizons in the study and diagnosis of age-related diseases.

Dremin Viktor, Marcinkevics Zbignevs, Zherebtsov Evgeny, Popov Alexey, Grabovskis Andris, Kronberga Hedviga, Geldnere Kristine, Doronin Alexander, Meglinski Igor, Bykov Alexander


Radiology Radiology

Deep-learning for predicting C-shaped canals in mandibular second molars on panoramic radiographs.

In Dento maxillo facial radiology

OBJECTIVE : The aim of this study was to evaluate the use of a convolutional neural network (CNN) system for predicting C-shaped canals in mandibular second molars on panoramic radiographs.

METHODS : Panoramic and cone beam CT (CBCT) images obtained from June 2018 to May 2020 were screened and 1020 patients were selected. Our dataset of 2040 sound mandibular second molars comprised 887 C-shaped canals and 1153 non-C-shaped canals. To confirm the presence of a C-shaped canal, CBCT images were analyzed by a radiologist and set as the gold standard. A CNN-based deep-learning model for predicting C-shaped canals was built using Xception. The training and test sets were set to 80 to 20%, respectively. Diagnostic performance was evaluated using accuracy, sensitivity, specificity, and precision. Receiver-operating characteristics (ROC) curves were drawn, and the area under the curve (AUC) values were calculated. Further, gradient-weighted class activation maps (Grad-CAM) were generated to localize the anatomy that contributed to the predictions.

RESULTS : The accuracy, sensitivity, specificity, and precision of the CNN model were 95.1, 92.7, 97.0, and 95.9%, respectively. Grad-CAM analysis showed that the CNN model mainly identified root canal shapes converging into the apex to predict the C-shaped canals, while the root furcation was predominantly used for predicting the non-C-shaped canals.

CONCLUSIONS : The deep-learning system had significant accuracy in predicting C-shaped canals of mandibular second molars on panoramic radiographs.

Jeon Su-Jin, Yun Jong-Pil, Yeom Han-Gyeol, Shin Woo-Sang, Lee Jong-Hyun, Jeong Seung-Hyun, Seo Min-Seock


C-shaped canal, Convolutional neural network, Deep learning, Diagnostic imaging, Panoramic radiograph

General General

Permutationally Restrained Diabatization by Machine Intelligence.

In Journal of chemical theory and computation

Simulations of electronically nonadiabatic processes may employ either the adiabatic or diabatic representation. Direct dynamics calculations are usually carried out in the adiabatic basis because the energy, force, and state coupling can be evaluated directly by many electronic structure methods. However, although its straightforwardness is appealing, direct dynamics is expensive when combined with quantitatively accurate electronic structure theories. This generates interest in analytically fitted surfaces to cut the expense, but the cuspidal ridges of the potentials and the singularities and vector nature of the couplings at high-dimensional, nonsymmetry-determined intersections in the adiabatic representation make accurate fitting almost impossible. This motivates using diabatic representations, where the surfaces are smooth and the couplings are also smooth and-importantly-scalar. In a recent previous work, we have developed a method called diabatization by deep neural network (DDNN) that takes advantage of the smoothness and nonuniqueness of diabatic bases to obtain them by machine learning. The diabatic potential energy matrices (DPEMs) learned by the DDNN method yield not only diabatic potential energy surfaces (PESs) and couplings in an analytic form useful for dynamics calculations, but also adiabatic surfaces and couplings in the adiabatic representation can be calculated inexpensively from the transformation. In the present work, we show how to extend the DDNN method to produce good approximations to global permutationally invariant adiabatic PESs simultaneously with DPEMs. The extended method is called permutationally restrained DDNN.

Shu Yinan, Varga Zoltan, Sampaio de Oliveira-Filho Antonio Gustavo, Truhlar Donald G


General General

A Role for Artificial Intelligence in the Classification of Craniofacial Anomalies.

In The Journal of craniofacial surgery

Development of an objective algorithm to diagnose and assess craniofacial conditions has the potential to facilitate early diagnosis, especially for care providers with limited craniofacial expertise. Deep learning, a branch of artificial intelligence, can automatically analyze and categorize disease without human assistance. Convolutional neural networks (CNN) have excelled in utilizing medical images to automatically classify disease. In this study, the authors developed CNN models to detect and classify non-syndromic craniosynostosis (CS) using 2D images. The authors created an annotated data set of labeled CS (normal, metopic, sagittal, and unicoronal) conditions using standard clinical photography from the image repository at our center. The authors extended this dataset set by adding photographic images of children with craniofacial conditions from the internet. A total of 1076 images were used in this study. The authors developed a CNN model using a pre-trained ResNet-50 model to classify the data as metopic, sagittal, and unicoronal. The testing accuracy for the CS ResNet50 model achieved an overall testing accuracy of 90.6%. The sensitivity and precision were: 100% and 100% for metopic, 93.3% and 100% for sagittal, and 66.7% and 100% for unicoronal, respectively. The CNN model performed with promising accuracy. These results support the idea that deep learning has a role in diagnosis of craniofacial conditions. Using standard 2D clinical photography, such systems can provide automated screening and detection of these conditions. In the future, ML may be applied to prediction and assessment of surgical outcomes, or as an open-source remote diagnostic resource.

Geisler Emily L, Agarwal Saloni, Hallac Rami R, Daescu Ovidiu, Kane Alex A