Estimation of human age using machine learning on panoramic radiographs for Brazilian patients.

This paper addresses a relevant problem in Forensic Sciences by integrating radiological techniques with advanced machine learning methodologies to create a non-invasive, efficient, and less examiner-dependent approach to age estimation. Our study includes a new dataset of 12,827 dental panoramic X-ray images representing the Brazilian population, covering an age range from 2.25 to 96.50 years. To analyze these exams, we employed a model adapted from InceptionV4, enhanced with data augmentation techniques. The proposed approach achieved robust and reliable results, with a Test Mean Absolute Error of 3.1 years and an R-squared value of 95.5%. Professional radiologists have validated that our model focuses on critical features for age assessment used in odontology, such as pulp chamber dimensions and stages of permanent teeth calcification. Importantly, the model also relies on anatomical information from the mandible, maxillary sinus, and vertebrae, which enables it to perform well even in edentulous cases. This study demonstrates the significant potential of machine learning to revolutionize age estimation in Forensic Science, offering a more accurate, efficient, and universally applicable solution.

Hearing recovery prediction and prognostic factors of idiopathic sudden sensorineural hearing loss: a retrospective analysis with a deep neural network model.

OBJECTIVE: Idiopathic Sudden Sensorineural Hearing Loss (ISSHL) is an otologic emergency, and an early prediction of prognosis may facilitate proper treatment. Therefore, we investigated the prognostic factors for predicting the recovery in patients with ISSHL treated with combined treatment method using machine learning models. METHODS: We retrospectively reviewed the medical records of 298 patients with ISSHL at a tertiary medical institution between January 2015 and September 2020. Fifty-two variables were analyzed to predict hearing recovery. Recovery was defined using Siegel's criteria, and the patients were categorized into recovery and non-recovery groups. Recovery was predicted by various machine learning models. In addition, the prognostic factors were analyzed using the difference in the loss function. RESULTS: There were significant differences in variables including age, hypertension, previous hearing loss, ear fullness, duration of hospital admission, initial hearing level of the affected and unaffected ears, and post-treatment hearing level between recovery and non-recovery groups. The deep neural network model showed the highest predictive performance (accuracy, 88.81%; area under the receiver operating characteristic curve, 0.9448). In addition, initial hearing level of affected and non-affected ear, post-treatment (2-weeks) hearing level of affected ear were significant factors for predicting the prognosis. CONCLUSION: The deep neural network model showed the highest predictive performance for recovery in patients with ISSHL. Some factors with prognostic value were identified. Further studies using a larger patient population are warranted. LEVEL OF EVIDENCE: Level 4.

Hearing recovery prediction and prognostic factors of idiopathic sudden sensorineural hearing loss: a retrospective analysis with a deep neural network model

Abstract Objective Idiopathic Sudden Sensorineural Hearing Loss (ISSHL) is an otologic emergency, and an early prediction of prognosis may facilitate proper treatment. Therefore, we investigated the prognostic factors for predicting the recovery in patients with ISSHL treated with combined treatment method using machine learning models. Methods We retrospectively reviewed the medical records of 298 patients with ISSHL at a tertiary medical institution between January 2015 and September 2020. Fifty-two variables were analyzed to predict hearing recovery. Recovery was defined using Siegel's criteria, and the patients were categorized into recovery and non-recovery groups. Recovery was predicted by various machine learning models. In addition, the prognostic factors were analyzed using the difference in the loss function. Results There were significant differences in variables including age, hypertension, previous hearing loss, ear fullness, duration of hospital admission, initial hearing level of the affected and unaffected ears, and post-treatment hearing level between recovery and non-recovery groups. The deep neural network model showed the highest predictive performance (accuracy, 88.81%; area under the receiver operating characteristic curve, 0.9448). In addition, initial hearing level of affected and non-affected ear, post-treatment (2-weeks) hearing level of affected ear were significant factors for predicting the prognosis. Conclusion The deep neural network model showed the highest predictive performance for recovery in patients with ISSHL. Some factors with prognostic value were identified. Further studies using a larger patient population are warranted. Level of evidence: Level 4.

NIFtHool: an informatics program for identification of NifH proteins using deep neural networks.

Atmospheric nitrogen fixation carried out by microorganisms has environmental and industrial importance, related to the increase of soil fertility and productivity. The present work proposes the development of a new high precision system that allows the recognition of amino acid sequences of the nitrogenase enzyme (NifH) as a promising way to improve the identification of diazotrophic bacteria. For this purpose, a database obtained from UniProt built a processed dataset formed by a set of 4911 and 4782 amino acid sequences of the NifH and non-NifH proteins respectively. Subsequently, the feature extraction was developed using two methodologies: (i) k-mers counting and (ii) embedding layers to obtain numerical vectors of the amino acid chains. Afterward, for the embedding layer, the data was crossed by an external trainable convolutional layer, which received a uniform matrix and applied convolution using filters to obtain the feature maps of the model. Finally, a deep neural network was used as the primary model to classify the amino acid sequences as NifH protein or not. Performance evaluation experiments were carried out, and the results revealed an accuracy of 96.4%, a sensitivity of 95.2%, and a specificity of 96.7%. Therefore, an amino acid sequence-based feature extraction method that uses a neural network to detect N-fixing organisms is proposed and implemented. NIFtHool is available from: https://nifthool.anvil.app/.

Accuracy of Deep Neural Network in Triaging Common Skin Diseases of Primary Care Attention.

Access to dermatological care can be challenging in certain regions of the world. The triage process is usually conducted by primary care physicians; however, they may not be able to diagnose and assign the correct referral and level of priority for different dermatosis. The present research aimed to test different deep neural networks to obtain the highest level of accuracy for the following: (1) diagnosing groups of dermatoses; (2) correct referrals; and (3) the level of priority given to the referral compared to dermatologists. Using 140,446 images from a teledermatology project, previously labeled with the clinical diagnosis, and their respective referrals, namely biopsy, in-person dermatologist visits or monitoring the case via teledermatology along with the general physician, 27 different scenarios of neural networks were derived, and the algorithm accuracies in classifying different dermatosis, according to the group of the diagnosis they belong to, were calculated. The most accurate algorithm was then tested for accuracy in diagnosis, referral, and level of priority given to 6,945 cases. The GoogLeNet architecture, trained with 24,000 images and 1,000 epochs, using weight random initialization and learning rates of 10-3 was found to be the most accurate network, showing an accuracy of 89.72% for diagnosis, 96.03% for referrals and 92.54% for priority level in 6,975 image testing. Our study population, however, was confined to individuals with chronic skin conditions and, therefore, it has limited value as a triage tool because it has not been tested for acute conditions. Deep neural networks are accurate in triaging, correct referral and prioritizing common chronic skin diseases related to primary care attention. They can also help health-care systems optimize patients' access to dermatologists.

Efficient Face Recognition System for Operating in Unconstrained Environments.

Facial recognition is fundamental for a wide variety of security systems operating in real-time applications. Recently, several deep neural networks algorithms have been developed to achieve state-of-the-art performance on this task. The present work was conceived due to the need for an efficient and low-cost processing system, so a real-time facial recognition system was proposed using a combination of deep learning algorithms like FaceNet and some traditional classifiers like SVM, KNN, and RF using moderate hardware to operate in an unconstrained environment. Generally, a facial recognition system involves two main tasks: face detection and recognition. The proposed scheme uses the YOLO-Face method for the face detection task which is a high-speed real-time detector based on YOLOv3, while, for the recognition stage, a combination of FaceNet with a supervised learning algorithm, such as the support vector machine (SVM), is proposed for classification. Extensive experiments on unconstrained datasets demonstrate that YOLO-Face provides better performance when the face under an analysis presents partial occlusion and pose variations; besides that, it can detect small faces. The face detector was able to achieve an accuracy of over 89.6% using the Honda/UCSD dataset which runs at 26 FPS with darknet-53 to VGA-resolution images for classification tasks. The experimental results have demonstrated that the FaceNet+SVM model was able to achieve an accuracy of 99.7% using the LFW dataset. On the same dataset, FaceNet+KNN and FaceNet+RF achieve 99.5% and 85.1%, respectively; on the other hand, the FaceNet was able to achieve 99.6%. Finally, the proposed system provides a recognition accuracy of 99.1% and 49 ms runtime when both the face detection and classifications stages operate together.

Predictive habitat suitability modeling of deep-sea framework-forming scleractinian corals in the Gulf of Mexico.

Framework-forming scleractinian (FFS) corals provide structurally complex habitats to support abundant and diverse benthic communities but are vulnerable to environmental changes and anthropogenic disturbances. Scientific modeling of suitable habitat provides important insights into the impact of the environmental conditions and fills the gap in the knowledge on habitat suitability. This study presents predictive habitat suitability modeling for deep-sea (depth > 50 m) FFS corals in the GoM. We first conducted a nonparametric estimate of the observed coral point process intensity as a function of each numeric environmental variable. Next, we performed species distribution modeling (SDM) using an assemble of four machine learning models - maximum entropy (ME), support vector machine (SVM), random forest (RF), and deep neural network (DNN). We found that most important variables controlling the coral distribution are super-dominant gravel and rock substrata, SW and SE aspects, slope steepness, salinity, depth, temperature, acidity, dissolved oxygen, and chlorophyll-a. Highly suitable habitats are predicted to be on the continental slope off Texas, Louisiana, and Mississippi and the shelf and slope of the West Florida Escarpment. All the four models have outstanding prediction performances with AUC values over 0.95. DNN model performs best (AUC = 0.987). The study contributes to coral habitat modeling research by presenting unique methods including nonparametric function of coral point process intensity, DNN and SVM models that have not been used in coral SDM, post-classification model assembling, and percentile approach to determine a threshold value for classifying a suitability score map into a binary map. Our findings would help support conservation prioritization, management and planning, and guide new field exploration.