J Chem Inf Model. 2024 Jun 3. doi: 10.1021/acs.jcim.4c00467. Online ahead of print.

ABSTRACT

Microbial rhodopsins (MRs) are a diverse and abundant family of photoactive membrane proteins that serve as model systems for biophysical techniques. Optogenetics utilizes genetic engineering to insert specialized proteins into specific neurons or brain regions, allowing for manipulation of their activity through light and enabling the mapping and control of specific brain areas in living organisms. The obstacle of optogenetics lies in the fact that light has a limited ability to penetrate biological tissues, particularly blue light in the visible spectrum. Despite this challenge, most optogenetic systems rely on blue light due to the scarcity of red-shifted opsins. Finding additional red-shifted rhodopsins would represent a major breakthrough in overcoming the challenge of limited light penetration in optogenetics. However, determining the wavelength absorption maxima for rhodopsins based on their protein sequence is a significant hurdle. Current experimental methods are time-consuming, while computational methods lack accuracy. The paper introduces a new computational approach called RhoMax that utilizes structure-based geometric deep learning to predict the absorption wavelength of rhodopsins solely based on their sequences. The method takes advantage of AlphaFold2 for accurate modeling of rhodopsin structures. Once trained on a balanced train set, RhoMax rapidly and precisely predicted the maximum absorption wavelength of more than half of the sequences in our test set with an accuracy of 0.03 eV. By leveraging computational methods for absorption maxima determination, we can drastically reduce the time needed for designing new red-shifted microbial rhodopsins, thereby facilitating advances in the field of optogenetics.

PMID:38829021 | DOI:10.1021/acs.jcim.4c00467

Nucleic Acids Res. 2024 Jun 3:gkae468. doi: 10.1093/nar/gkae468. Online ahead of print.

ABSTRACT

The computational design of synthetic DNA sequences with designer in vivo properties is gaining traction in the field of synthetic genomics. We propose here a computational method which combines a kinetic Monte Carlo framework with a deep mutational screening based on deep learning predictions. We apply our method to build regular nucleosome arrays with tailored nucleosomal repeat lengths (NRL) in yeast. Our design was validated in vivo by successfully engineering and integrating thousands of kilobases long tandem arrays of computationally optimized sequences which could accommodate NRLs much larger than the yeast natural NRL (namely 197 and 237 bp, compared to the natural NRL of ∼165 bp). RNA-seq results show that transcription of the arrays can occur but is not driven by the NRL. The computational method proposed here delineates the key sequence rules for nucleosome positioning in yeast and should be easily applicable to other sequence properties and other genomes.

PMID:38828788 | DOI:10.1093/nar/gkae468

Front Plant Sci. 2024 May 17;15:1382416. doi: 10.3389/fpls.2024.1382416. eCollection 2024.

ABSTRACT

Tomato is one of the most popular and most important food crops consumed globally. The quality and quantity of yield by tomato plants are affected by the impact made by various kinds of diseases. Therefore, it is essential to identify these diseases early so that it is possible to reduce the occurrences and effect of the diseases on tomato plants to improve the overall crop yield and to support the farmers. In the past, many research works have been carried out by applying the machine learning techniques to segment and classify the tomato leaf images. However, the existing machine learning-based classifiers are not able to detect the new types of diseases more accurately. On the other hand, deep learning-based classifiers with the support of swarm intelligence-based optimization techniques are able to enhance the classification accuracy, leading to the more effective and accurate detection of leaf diseases. This research paper proposes a new method for the accurate classification of tomato leaf diseases by harnessing the power of an ensemble model in a sample dataset of tomato plants, containing images pertaining to nine different types of leaf diseases. This research introduces an ensemble model with an exponential moving average function with temporal constraints and an enhanced weighted gradient optimizer that is integrated into fine-tuned Visual Geometry Group-16 (VGG-16) and Neural Architecture Search Network (NASNet) mobile training methods for providing improved learning and classification accuracy. The dataset used for the research consists of 10,000 tomato leaf images categorized into nine classes for training and validating the model and an additional 1,000 images reserved for testing the model. The results have been analyzed thoroughly and benchmarked with existing performance metrics, thus proving that the proposed approach gives better performance in terms of accuracy, loss, precision, recall, receiver operating characteristic curve, and F1-score with values of 98.7%, 4%, 97.9%, 98.6%, 99.97%, and 98.7%, respectively.

PMID:38828218 | PMC:PMC11140105 | DOI:10.3389/fpls.2024.1382416

Biol Imaging. 2024 May 9;4:e7. doi: 10.1017/S2633903X24000060. eCollection 2024.

ABSTRACT

Cryogenic electron tomography (cryoET) is capable of determining in situ biological structures of molecular complexes at near-atomic resolution by averaging half a million subtomograms. While abundant complexes/particles are often clustered in arrays, precisely locating and seamlessly averaging such particles across many tomograms present major challenges. Here, we developed TomoNet, a software package with a modern graphical user interface to carry out the entire pipeline of cryoET and subtomogram averaging to achieve high resolution. TomoNet features built-in automatic particle picking and three-dimensional (3D) classification functions and integrates commonly used packages to streamline high-resolution subtomogram averaging for structures in 1D, 2D, or 3D arrays. Automatic particle picking is accomplished in two complementary ways: one based on template matching and the other using deep learning. TomoNet's hierarchical file organization and visual display facilitate efficient data management as required for large cryoET datasets. Applications of TomoNet to three types of datasets demonstrate its capability of efficient and accurate particle picking on flexible and imperfect lattices to obtain high-resolution 3D biological structures: virus-like particles, bacterial surface layers within cellular lamellae, and membranes decorated with nuclear egress protein complexes. These results demonstrate TomoNet's potential for broad applications to various cryoET projects targeting high-resolution in situ structures.

PMID:38828212 | PMC:PMC11140495 | DOI:10.1017/S2633903X24000060

Eur J Radiol Open. 2024 May 24;12:100570. doi: 10.1016/j.ejro.2024.100570. eCollection 2024 Jun.

ABSTRACT

PURPOSE: Super-resolution deep-learning-based reconstruction: SR-DLR is a newly developed and clinically available deep-learning-based image reconstruction method that can improve the spatial resolution of CT images. The image quality of the output from non-linear image reconstructions, such as DLR, is known to vary depending on the structure of the object being scanned, and a simple phantom cannot explicitly evaluate the clinical performance of SR-DLR. This study aims to accurately investigate the quality of the images reconstructed by SR-DLR by utilizing a structured phantom that simulates the human anatomy in coronary CT angiography.

METHODS: The structural phantom had ribs and vertebrae made of plaster, a left ventricle filled with dilute contrast medium, a coronary artery with simulated stenosis, and an implanted stent graft. By scanning the structured phantom, we evaluated noise and spatial resolution on the images reconstructed with SR-DLR and conventional reconstructions.

RESULTS: The spatial resolution of SR-DLR was higher than conventional reconstructions; the 10 % modulation transfer function of hybrid IR (HIR), DLR, and SR-DLR were 0.792-, 0.976-, and 1.379 cycle/mm, respectively. At the same time, image noise was lowest (HIR: 21.1-, DLR: 19.0-, and SR-DLR: 13.1 HU). SR-DLR could accurately assess coronary artery stenosis and the lumen of the implanted stent graft.

CONCLUSIONS: SR-DLR can obtain CT images with high spatial resolution and lower noise without special CT equipments, and will help diagnose coronary artery disease in CCTA and other CT examinations that require high spatial resolution.

PMID:38828096 | PMC:PMC11140562 | DOI:10.1016/j.ejro.2024.100570

J Med Phys. 2024 Jan-Mar;49(1):41-48. doi: 10.4103/jmp.jmp_106_23. Epub 2024 Mar 30.

ABSTRACT

PURPOSE: Errors in the identification of true patients in a health-care facility may result in the wrong dose or dosage being given to the wrong patient at the wrong site during radiotherapy sessions, radiopharmaceutical administration, radiological scans, etc. The aim of this article is to reduce the error in the identification of correct patients by implementation of the Python deep learning-based real-time patient identification program.

MATERIALS AND METHODS: The authors utilized and installed Anaconda Prompt (miniconda 3), Python (version 3.9.12), and Visual Studio Code (version 1.71.0) for the design of the patient identification program. In the field of view, the area of interest is merely face detection. The overall performance of the developed program is accomplished over three steps, namely image data collection, data transfer, and data analysis, respectively. The patient identification tool was developed using the OpenCV library for face recognition.

RESULTS: This program provides real-time patient identification information, together with the other preset parameters such as disease site, with a precision of 0.92%, recall rate of 0.80%, and specificity of 0.90%. Furthermore, the accuracy of the program was found to be 0.84%. The output of the in-house developed program as "Unknown" is provided if a patient's relative or an unknown person is found in restricted region.

INTERPRETATION AND CONCLUSIONS: This Python-based program is beneficial for confirming the patient's identity, without manual interventions, just before therapy, administering medications, and starting other medical procedures, among other things, to prevent unintended medical and health-related complications that may arise as a result of misidentification.

PMID:38828072 | PMC:PMC11141754 | DOI:10.4103/jmp.jmp_106_23

J Med Phys. 2024 Jan-Mar;49(1):33-40. doi: 10.4103/jmp.jmp_122_23. Epub 2024 Mar 30.

ABSTRACT

PURPOSE: This study aimed to develop a deep learning model for the prediction of V20 (the volume of the lung parenchyma that received ≥20 Gy) during intensity-modulated radiation therapy using chest X-ray images.

METHODS: The study utilized 91 chest X-ray images of patients with lung cancer acquired routinely during the admission workup. The prescription dose for the planning target volume was 60 Gy in 30 fractions. A convolutional neural network-based regression model was developed to predict V20. To evaluate model performance, the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) were calculated with conducting a four-fold cross-validation method. The patient characteristics of the eligible data were treatment period (2018-2022) and V20 (19.3%; 4.9%-30.7%).

RESULTS: The predictive results of the developed model for V20 were 0.16, 5.4%, and 4.5% for the R2, RMSE, and MAE, respectively. The median error was -1.8% (range, -13.0% to 9.2%). The Pearson correlation coefficient between the calculated and predicted V20 values was 0.40. As a binary classifier with V20 <20%, the model showed a sensitivity of 75.0%, specificity of 82.6%, diagnostic accuracy of 80.6%, and area under the receiver operator characteristic curve of 0.79.

CONCLUSIONS: The proposed deep learning chest X-ray model can predict V20 and play an important role in the early determination of patient treatment strategies.

PMID:38828071 | PMC:PMC11141742 | DOI:10.4103/jmp.jmp_122_23

J Med Imaging (Bellingham). 2024 May;11(3):034504. doi: 10.1117/1.JMI.11.3.034504. Epub 2024 May 30.

ABSTRACT

PURPOSE: Accurate segmentation of the endometrium in ultrasound images is essential for gynecological diagnostics and treatment planning. Manual segmentation methods are time-consuming and subjective, prompting the exploration of automated solutions. We introduce "segment anything with inception module" (SAIM), a specialized adaptation of the segment anything model, tailored specifically for the segmentation of endometrium structures in ultrasound images.

APPROACH: SAIM incorporates enhancements to the image encoder structure and integrates point prompts to guide the segmentation process. We utilized ultrasound images from patients undergoing hysteroscopic surgery in the gynecological department to train and evaluate the model.

RESULTS: Our study demonstrates SAIM's superior segmentation performance through quantitative and qualitative evaluations, surpassing existing automated methods. SAIM achieves a dice similarity coefficient of 76.31% and an intersection over union score of 63.71%, outperforming traditional task-specific deep learning models and other SAM-based foundation models.

CONCLUSIONS: The proposed SAIM achieves high segmentation accuracy, providing high diagnostic precision and efficiency. Furthermore, it is potentially an efficient tool for junior medical professionals in education and diagnosis.

PMID:38827779 | PMC:PMC11137375 | DOI:10.1117/1.JMI.11.3.034504

J Med Imaging (Bellingham). 2024 May;11(3):036002. doi: 10.1117/1.JMI.11.3.036002. Epub 2024 May 30.

ABSTRACT

PURPOSE: Early detection of cancer is crucial for lung cancer patients, as it determines disease prognosis. Lung cancer typically starts as bronchial lesions along the airway walls. Recent research has indicated that narrow-band imaging (NBI) bronchoscopy enables more effective bronchial lesion detection than other bronchoscopic modalities. Unfortunately, NBI video can be hard to interpret because physicians currently are forced to perform a time-consuming subjective visual search to detect bronchial lesions in a long airway-exam video. As a result, NBI bronchoscopy is not regularly used in practice. To alleviate this problem, we propose an automatic two-stage real-time method for bronchial lesion detection in NBI video and perform a first-of-its-kind pilot study of the method using NBI airway exam video collected at our institution.

APPROACH: Given a patient's NBI video, the first method stage entails a deep-learning-based object detection network coupled with a multiframe abnormality measure to locate candidate lesions on each video frame. The second method stage then draws upon a Siamese network and a Kalman filter to track candidate lesions over multiple frames to arrive at final lesion decisions.

RESULTS: Tests drawing on 23 patient NBI airway exam videos indicate that the method can process an incoming video stream at a real-time frame rate, thereby making the method viable for real-time inspection during a live bronchoscopic airway exam. Furthermore, our studies showed a 93% sensitivity and 86% specificity for lesion detection; this compares favorably to a sensitivity and specificity of 80% and 84% achieved over a series of recent pooled clinical studies using the current time-consuming subjective clinical approach.

CONCLUSION: The method shows potential for robust lesion detection in NBI video at a real-time frame rate. Therefore, it could help enable more common use of NBI bronchoscopy for bronchial lesion detection.

PMID:38827776 | PMC:PMC11138083 | DOI:10.1117/1.JMI.11.3.036002

Ophthalmol Sci. 2024 Jan 26;4(5):100477. doi: 10.1016/j.xops.2024.100477. eCollection 2024 Sep-Oct.

ABSTRACT

PURPOSE: To gain an understanding of data labeling requirements to train deep learning models for measurement of geographic atrophy (GA) with fundus autofluorescence (FAF) images.

DESIGN: Evaluation of artificial intelligence (AI) algorithms.

SUBJECTS: The Age-Related Eye Disease Study 2 (AREDS2) images were used for training and cross-validation, and GA clinical trial images were used for testing.

METHODS: Training data consisted of 2 sets of FAF images; 1 with area measurements only and no indication of GA location (Weakly labeled) and the second with GA segmentation masks (Strongly labeled).

MAIN OUTCOME MEASURES: Bland-Altman plots and scatter plots were used to compare GA area measurement between ground truth and AI measurements. The Dice coefficient was used to compare accuracy of segmentation of the Strong model.

RESULTS: In the cross-validation AREDS2 data set (n = 601), the mean (standard deviation [SD]) area of GA measured by human grader, Weakly labeled AI model, and Strongly labeled AI model was 6.65 (6.3) mm2, 6.83 (6.29) mm2, and 6.58 (6.24) mm2, respectively. The mean difference between ground truth and AI was 0.18 mm2 (95% confidence interval, [CI], -7.57 to 7.92) for the Weakly labeled model and -0.07 mm2 (95% CI, -1.61 to 1.47) for the Strongly labeled model. With GlaxoSmithKline testing data (n = 156), the mean (SD) GA area was 9.79 (5.6) mm2, 8.82 (4.61) mm2, and 9.55 (5.66) mm2 for human grader, Strongly labeled AI model, and Weakly labeled AI model, respectively. The mean difference between ground truth and AI for the 2 models was -0.97 mm2 (95% CI, -4.36 to 2.41) and -0.24 mm2 (95% CI, -4.98 to 4.49), respectively. The Dice coefficient was 0.99 for intergrader agreement, 0.89 for the cross-validation data, and 0.92 for the testing data.

CONCLUSIONS: Deep learning models can achieve reasonable accuracy even with Weakly labeled data. Training methods that integrate large volumes of Weakly labeled images with small number of Strongly labeled images offer a promising solution to overcome the burden of cost and time for data labeling.

FINANCIAL DISCLOSURES: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

PMID:38827491 | PMC:PMC11141255 | DOI:10.1016/j.xops.2024.100477

ArXiv [Preprint]. 2024 May 25:arXiv:2305.14749v5.

ABSTRACT

Computational RNA design tasks are often posed as inverse problems, where sequences are designed based on adopting a single desired secondary structure without considering 3D geometry and conformational diversity. We introduce gRNAde, a geometric RNA design pipeline operating on 3D RNA backbones to design sequences that explicitly account for structure and dynamics. Under the hood, gRNAde is a multi-state Graph Neural Network that generates candidate RNA sequences conditioned on one or more 3D backbone structures where the identities of the bases are unknown. On a single-state fixed backbone re-design benchmark of 14 RNA structures from the PDB identified by Das et al. [2010], gRNAde obtains higher native sequence recovery rates (56% on average) compared to Rosetta (45% on average), taking under a second to produce designs compared to the reported hours for Rosetta. We further demonstrate the utility of gRNAde on a new benchmark of multi-state design for structurally flexible RNAs, as well as zero-shot ranking of mutational fitness landscapes in a retrospective analysis of a recent RNA polymerase ribozyme structure. Open source code: https://github.com/chaitjo/geometric-rna-design.

PMID:38827456 | PMC:PMC11142323

ArXiv [Preprint]. 2024 May 23:arXiv:2405.14108v1.

ABSTRACT

The effects of ligand binding on protein structures and their in vivo functions carry numerous implications for modern biomedical research and biotechnology development efforts such as drug discovery. Although several deep learning (DL) methods and benchmarks designed for protein-ligand docking have recently been introduced, to date no prior works have systematically studied the behavior of docking methods within the practical context of (1) predicted (apo) protein structures, (2) multiple ligands concurrently binding to a given target protein, and (3) having no prior knowledge of binding pockets. To enable a deeper understanding of docking methods' real-world utility, we introduce PoseBench, the first comprehensive benchmark for practical protein-ligand docking. PoseBench enables researchers to rigorously and systematically evaluate DL docking methods for apo-to-holo protein-ligand docking and protein-ligand structure generation using both single and multi-ligand benchmark datasets, the latter of which we introduce for the first time to the DL community. Empirically, using PoseBench, we find that all recent DL docking methods but one fail to generalize to multi-ligand protein targets and also that template-based docking algorithms perform equally well or better for multi-ligand docking as recent single-ligand DL docking methods, suggesting areas of improvement for future work. Code, data, tutorials, and benchmark results are available at https://github.com/BioinfoMachineLearning/PoseBench.

PMID:38827451 | PMC:PMC11142318

Spectrochim Acta A Mol Biomol Spectrosc. 2024 May 28;319:124546. doi: 10.1016/j.saa.2024.124546. Online ahead of print.

ABSTRACT

Mid-infrared (MIR) spectroscopy can characterize the content and structural changes of macromolecular components in different breast tissues, which can be used for feature extraction and model training by machine learning to achieve accurate classification and recognition of different breast tissues. In parallel, the one-dimensional convolutional neural network (1D-CNN) stands out in the field of deep learning for its ability to efficiently process sequential data, such as spectroscopic signals. In this study, MIR spectra of breast tissue were collected in situ by coupling the self-developed MIR hollow optical fiber attenuated total reflection (HOF-ATR) probe with a Fourier transform infrared spectroscopy (FTIR) spectrometer. Staging analysis was conducted on the changes in macromolecular content and structure in breast cancer tissues. For the first time, a trinary classification model was established based on 1D-CNN for recognizing normal, paracancerous and cancerous tissues. The final predication results reveal that the 1D-CNN model based on baseline correction (BC) and data augmentation yields more precise classification results, with a total accuracy of 95.09%, exhibiting superior discrimination ability than machine learning models of SVM-DA (90.00%), SVR (88.89%), PCA-FDA (67.78%) and PCA-KNN (70.00%). The experimental results suggest that the application of 1D-CNN enables accurate classification and recognition of different breast tissues, which can be considered as a precise, efficient and intelligent novel method for breast cancer diagnosis.

PMID:38824755 | DOI:10.1016/j.saa.2024.124546

Comput Med Imaging Graph. 2024 May 28;116:102405. doi: 10.1016/j.compmedimag.2024.102405. Online ahead of print.

ABSTRACT

Over the past decade, deep-learning (DL) algorithms have become a promising tool to aid clinicians in identifying fetal head standard planes (FHSPs) during ultrasound (US) examination. However, the adoption of these algorithms in clinical settings is still hindered by the lack of large annotated datasets. To overcome this barrier, we introduce FetalBrainAwareNet, an innovative framework designed to synthesize anatomically accurate images of FHSPs. FetalBrainAwareNet introduces a cutting-edge approach that utilizes class activation maps as a prior in its conditional adversarial training process. This approach fosters the presence of the specific anatomical landmarks in the synthesized images. Additionally, we investigate specialized regularization terms within the adversarial training loss function to control the morphology of the fetal skull and foster the differentiation between the standard planes, ensuring that the synthetic images faithfully represent real US scans in both structure and overall appearance. The versatility of our FetalBrainAwareNet framework is highlighted by its ability to generate high-quality images of three predominant FHSPs using a singular, integrated framework. Quantitative (Fréchet inception distance of 88.52) and qualitative (t-SNE) results suggest that our framework generates US images with greater variability compared to state-of-the-art methods. By using the synthetic images generated with our framework, we increase the accuracy of FHSP classifiers by 3.2% compared to training the same classifiers solely with real acquisitions. These achievements suggest that using our synthetic images to increase the training set could provide benefits to enhance the performance of DL algorithms for FHSPs classification that could be integrated in real clinical scenarios.

PMID:38824716 | DOI:10.1016/j.compmedimag.2024.102405

Comput Med Imaging Graph. 2024 May 28;116:102406. doi: 10.1016/j.compmedimag.2024.102406. Online ahead of print.

ABSTRACT

Lack of data is one of the biggest hurdles for rare disease research using deep learning. Due to the lack of rare-disease images and annotations, training a robust network for automatic rare-disease image segmentation is very challenging. To address this challenge, few-shot domain adaptation (FSDA) has emerged as a practical research direction, aiming to leverage a limited number of annotated images from a target domain to facilitate adaptation of models trained on other large datasets in a source domain. In this paper, we present a novel prototype-based feature mapping network (PFMNet) designed for FSDA in medical image segmentation. PFMNet adopts an encoder-decoder structure for segmentation, with the prototype-based feature mapping (PFM) module positioned at the bottom of the encoder-decoder structure. The PFM module transforms high-level features from the target domain into the source domain-like features that are more easily comprehensible by the decoder. By leveraging these source domain-like features, the decoder can effectively perform few-shot segmentation in the target domain and generate accurate segmentation masks. We evaluate the performance of PFMNet through experiments on three typical yet challenging few-shot medical image segmentation tasks: cross-center optic disc/cup segmentation, cross-center polyp segmentation, and cross-modality cardiac structure segmentation. We consider four different settings: 5-shot, 10-shot, 15-shot, and 20-shot. The experimental results substantiate the efficacy of our proposed approach for few-shot domain adaptation in medical image segmentation.

PMID:38824715 | DOI:10.1016/j.compmedimag.2024.102406

Ultrasound Med Biol. 2024 Jun 1:S0301-5629(24)00206-0. doi: 10.1016/j.ultrasmedbio.2024.05.001. Online ahead of print.

ABSTRACT

OBJECTIVE: Although there are methods to identify regions of interest (ROIs) from echocardiographic images of myocardial tissue, they are often time-consuming and difficult to create when image quality is poor. Further, while myocardial strain from ultrasound (US) images can be estimated, US alone cannot obtain functional information, such as oxygen saturation (sO2). Photoacoustic (PA) imaging, however, can be used to quantify sO2 levels within tissue non-invasively.

METHODS: Here, we leverage deep learning methods to improve segmentation of the anterior wall of the left ventricle and apply both strain and oxygen saturation analysis via segmentation of murine US and PA images.

RESULTS: Data revealed that training on US/PA images using a U-Net deep neural network can be used to create reproducible ROIs of the anterior wall of the left ventricle in a murine image dataset. Accuracy and Dice score metrics were used to evaluate performance of the neural network on each image type. We report an accuracy of 97.3% and Dice score of 0.84 for ultrasound, 95.6% and 0.73 for photoacoustic, and 96.5% and 0.81 for combined ultrasound and photoacoustic images.

CONCLUSION: Rapid segmentation via such methods can assist in quantification of strain and oxygenation.

PMID:38825556 | DOI:10.1016/j.ultrasmedbio.2024.05.001

J Tissue Viability. 2024 May 22:S0965-206X(24)00071-8. doi: 10.1016/j.jtv.2024.05.009. Online ahead of print.

ABSTRACT

BACKGROUND: The development of models using deep learning (DL) to assess pressure injuries from wound images has recently gained attention. Creating enough supervised data is important for improving performance but is time-consuming. Therefore, the development of models that can achieve high performance with limited supervised data is desirable.

MATERIALS AND METHODS: This retrospective observational study utilized DL and included patients who received medical examinations for sacral pressure injuries between February 2017 and December 2021. Images were labeled according to the DESIGN-R® classification. Three artificial intelligence (AI) models for assessing pressure injury depth were created with a convolutional neural network (Categorical, Binary, and Combined classification models) and performance was compared among the models.

RESULTS: A set of 414 pressure injury images in five depth stages (d0 to D4) were analyzed. The Combined classification model showed superior performance (F1-score, 0.868). The Categorical classification model frequently misclassified d1 and d2 as d0 (d0 Precision, 0.503), but showed high performance for D3 and D4 (F1-score, 0.986 and 0.966, respectively). The Binary classification model showed high performance in differentiating between d0 and d1-D4 (F1-score, 0.895); however, performance decreased with increasing number of evaluation steps.

CONCLUSION: The Combined classification model displayed superior performance without increasing the supervised data, which can be attributed to use of the high-performance Binary classification model for initial d0 evaluation and subsequent use of the Categorical classification model with fewer evaluation steps. Understanding the unique characteristics of classification methods and deploying them appropriately can enhance AI model performance.

PMID:38825443 | DOI:10.1016/j.jtv.2024.05.009

Biomed Eng Online. 2024 Jun 1;23(1):50. doi: 10.1186/s12938-024-01244-w.

ABSTRACT

BACKGROUND: Over 60% of epilepsy patients globally are children, whose early diagnosis and treatment are critical for their development and can substantially reduce the disease's burden on both families and society. Numerous algorithms for automated epilepsy detection from EEGs have been proposed. Yet, the occurrence of epileptic seizures during an EEG exam cannot always be guaranteed in clinical practice. Models that exclusively use seizure EEGs for detection risk artificially enhanced performance metrics. Therefore, there is a pressing need for a universally applicable model that can perform automatic epilepsy detection in a variety of complex real-world scenarios.

METHOD: To address this problem, we have devised a novel technique employing a temporal convolutional neural network with self-attention (TCN-SA). Our model comprises two primary components: a TCN for extracting time-variant features from EEG signals, followed by a self-attention (SA) layer that assigns importance to these features. By focusing on key features, our model achieves heightened classification accuracy for epilepsy detection.

RESULTS: The efficacy of our model was validated on a pediatric epilepsy dataset we collected and on the Bonn dataset, attaining accuracies of 95.50% on our dataset, and 97.37% (A v. E), and 93.50% (B vs E), respectively. When compared with other deep learning architectures (temporal convolutional neural network, self-attention network, and standardized convolutional neural network) using the same datasets, our TCN-SA model demonstrated superior performance in the automated detection of epilepsy.

CONCLUSION: The proven effectiveness of the TCN-SA approach substantiates its potential as a valuable tool for the automated detection of epilepsy, offering significant benefits in diverse and complex real-world clinical settings.

PMID:38824547 | DOI:10.1186/s12938-024-01244-w

BMC Infect Dis. 2024 Jun 1;24(1):551. doi: 10.1186/s12879-024-09428-4.

ABSTRACT

BACKGROUND: Leishmaniasis, an illness caused by protozoa, accounts for a substantial number of human fatalities globally, thereby emerging as one of the most fatal parasitic diseases. The conventional methods employed for detecting the Leishmania parasite through microscopy are not only time-consuming but also susceptible to errors. Therefore, the main objective of this study is to develop a model based on deep learning, a subfield of artificial intelligence, that could facilitate automated diagnosis of leishmaniasis.

METHODS: In this research, we introduce LeishFuNet, a deep learning framework designed for detecting Leishmania parasites in microscopic images. To enhance the performance of our model through same-domain transfer learning, we initially train four distinct models: VGG19, ResNet50, MobileNetV2, and DenseNet 169 on a dataset related to another infectious disease, COVID-19. These trained models are then utilized as new pre-trained models and fine-tuned on a set of 292 self-collected high-resolution microscopic images, consisting of 138 positive cases and 154 negative cases. The final prediction is generated through the fusion of information analyzed by these pre-trained models. Grad-CAM, an explainable artificial intelligence technique, is implemented to demonstrate the model's interpretability.

RESULTS: The final results of utilizing our model for detecting amastigotes in microscopic images are as follows: accuracy of 98.95 1.4%, specificity of 98 2.67%, sensitivity of 100%, precision of 97.91 2.77%, F1-score of 98.92 1.43%, and Area Under Receiver Operating Characteristic Curve of 99 1.33.

CONCLUSION: The newly devised system is precise, swift, user-friendly, and economical, thus indicating the potential of deep learning as a substitute for the prevailing leishmanial diagnostic techniques.

PMID:38824500 | DOI:10.1186/s12879-024-09428-4

Opt Lett. 2024 Jun 1;49(11):3138-3141. doi: 10.1364/OL.525115.

ABSTRACT

Visible light communication (VLC) technology with rich spectrum resources is thought of as an essential component in the future ubiquitous communication networks. Accurately monitoring its transmission impairments is important for improving the stability of high-speed communication networks. Existing research on intelligently monitoring the signal-to-noise ratio (SNR) performance of VLC focuses primarily on the application of neural networks but neglects the physical nature of communication systems. In this work, we propose an intelligent SNR estimation scheme for VLC systems, which is based on the symmetry of constellation diagrams with classical deep learning frameworks. In order to increase the accuracy of the SNR estimation scheme, we introduce two data augmentation methods (DA): point normalization and quadrant normalization. The results of extensive simulations demonstrate that the proposed point normalization method is capable of improving accuracy by about 5, 10, 14, and 26%, respectively, for 16-, 64-, 256-, and 1024-quadrature amplitude modulation compared with the same network frameworks without DA. The effect of accuracy improvement can be further superimposed with traditional DA methods. Additionally, the extensive number of constellation points (e.g., 32, 64, 128, 256, 512, 1024, and 2048) on the accuracy of SNR estimation is also investigated.

PMID:38824347 | DOI:10.1364/OL.525115