Mod Pathol. 2024 Jul 16:100563. doi: 10.1016/j.modpat.2024.100563. Online ahead of print.

ABSTRACT

The biopsy Gleason score is an important prognostic marker for prostate cancer patients. It is, however, subject to substantial variability among pathologists. Artificial intelligence (AI)-based algorithms employing deep learning have shown their ability to match pathologists' performance in assigning Gleason scores, with the potential to enhance pathologists' grading accuracy. The performance of Gleason AI algorithms in research is mostly reported on common benchmark datasets or within public challenges. In contrast, many commercial algorithms are evaluated in clinical studies, for which data is not publicly released. As commercial AI vendors typically do not publish performance on public benchmarks, comparison between research and commercial AI is difficult. The aim of this study is to evaluate and compare the performance of top-ranked public and commercial algorithms using real-world data. We curated a diverse dataset of whole-slide prostate biopsy images through crowdsourcing, containing images with a range of Gleason scores and from diverse sources. Predictions were obtained from five top-ranked public algorithms from the PANDA challenge and from two commercial Gleason grading algorithms. Additionally, ten pathologists evaluated the data set in a reader study. Overall, the pairwise quadratic weighted kappa among pathologists ranged from 0.777 to 0.916. Both public and commercial algorithms showed high agreement with pathologists, with quadratic kappa ranging from 0.617 to 0.900. Commercial algorithms performed on par or outperformed top public algorithms.

PMID:39025402 | DOI:10.1016/j.modpat.2024.100563

Surv Ophthalmol. 2024 Jul 16:S0039-6257(24)00081-X. doi: 10.1016/j.survophthal.2024.07.005. Online ahead of print.

ABSTRACT

Meibomian gland dysfunction (MGD) is increasingly recognized as a critical contributor to evaporative dry eye, significantly impacting visual quality. With a global prevalence estimated at 35.8%, it presents substantial challenges for clinicians. Conventional manual evaluation techniques for MGD face limitations characterized by inefficiencies, high subjectivity, limited big data processing capabilities, and a dearth of quantitative analytical tools. With rapidly advancing artificial intelligence (AI) technique revolutionizing ophthalmology, studies are now leveraging sophisticated AI methodologies, including computer vision, unsupervised learning, and supervised learning, to facilitate comprehensive analyses of meibomian gland (MG) evaluations. These evaluations employ various techniques, including slit lamp examination, infrared imaging, confocal microscopy, optical coherence tomography. This paradigm shift promises enhanced accuracy and consistency in disease evaluation and severity classification. While AI has achieved preliminary strides in meibomian gland evaluation, ongoing advancements in system development and clinical validation are imperative. We review the evolution of MG evaluation, juxtaposes AI-driven methods with traditional approaches, elucidates the specific roles of diverse AI technologies, and explores their practical applications using various evaluation techniques. Moreover, we delve into critical considerations for the clinical deployment of AI technologies and envisages future prospects, providing novel insights into MG evaluation and fostering technological and clinical progress in this arena.

PMID:39025239 | DOI:10.1016/j.survophthal.2024.07.005

Phys Med Biol. 2024 Jul 18. doi: 10.1088/1361-6560/ad6526. Online ahead of print.

ABSTRACT


Online adaptive radiation therapy requires fast and automated contouring of daily scans for treatment plan re-optimization. However, automated contouring is imperfect and introduces contour uncertainties. This work aims at developing and comparing robust optimization strategies accounting for such uncertainties.

Approach: 
A deep-learning method was used to predict the uncertainty of deformable image registration, and to generate a finite set of daily contour samples. Ten optimization strategies were compared: two baseline methods, five methods that convert contour samples into voxel-wise probabilities, and three methods accounting explicitly for contour samples as scenarios in robust optimization. Target coverage and organ-at-risk (OAR) sparing were evaluated robustly for simplified proton therapy plans for five head-and-neck cancer patients.

Results: 
We found that explicitly including target contour uncertainty in robust optimization provides robust target coverage with better OAR sparing than the baseline methods, without increasing the optimization time. Although OAR doses first increased when increasing target robustness, this effect could be prevented by additionally including robustness to OAR contour uncertainty. Compared to the probability-based methods, the scenario-based methods spared the OARs more, but increased integral dose and required more computation time.

Significance: 
This work proposed efficient and beneficial strategies to mitigate contour uncertainty in treatment plan optimization. This facilitates the adoption of automatic contouring in online adaptive radiation therapy and, more generally, enables mitigation also of other sources of contour uncertainty in treatment planning.

PMID:39025113 | DOI:10.1088/1361-6560/ad6526

Med Image Anal. 2024 Jul 11;97:103269. doi: 10.1016/j.media.2024.103269. Online ahead of print.

ABSTRACT

Lesion volume is an important predictor for prognosis in breast cancer. However, it is currently impossible to compute lesion volumes accurately from digital mammography data, which is the most popular and readily available imaging modality for breast cancer. We make a step towards a more accurate lesion volume measurement on digital mammograms by developing a model that allows to estimate lesion volumes on processed mammogram. Processed mammograms are the images routinely used by radiologists in clinical practice as well as in breast cancer screening and are available in medical centers. Processed mammograms are obtained from raw mammograms, which are the X-ray data coming directly from the scanner, by applying certain vendor-specific non-linear transformations. At the core of our volume estimation method is a physics-based algorithm for measuring lesion volumes on raw mammograms. We subsequently extend this algorithm to processed mammograms via a deep learning image-to-image translation model that produces synthetic raw mammograms from processed mammograms in a multi-vendor setting. We assess the reliability and validity of our method using a dataset of 1778 mammograms with an annotated mass. Firstly, we investigate the correlations between lesion volumes computed from mediolateral oblique and craniocaudal views, with a resulting Pearson correlation of 0.93 [95% confidence interval (CI) 0.92 - 0.93]. Secondly, we compare the resulting lesion volumes from true and synthetic raw data, with a resulting Pearson correlation of 0.998 [95%CI 0.998 - 0.998] . Finally, for a subset of 100 mammograms with a malignant mass and concurrent MRI examination available, we analyze the agreement between lesion volume on mammography and MRI, resulting in an intraclass correlation coefficient of 0.81 [95%CI 0.73 - 0.87] for consistency and 0.78 [95%CI 0.66 - 0.86] for absolute agreement. In conclusion, we developed an algorithm to measure mammographic lesion volume that reached excellent reliability and good validity, when using MRI as ground truth. The algorithm may play a role in lesion characterization and breast cancer prognostication on mammograms.

PMID:39024973 | DOI:10.1016/j.media.2024.103269

Med Image Anal. 2024 Jul 10;97:103272. doi: 10.1016/j.media.2024.103272. Online ahead of print.

ABSTRACT

Landmark detection is a crucial task in medical image analysis, with applications across various fields. However, current methods struggle to accurately locate landmarks in medical images with blurred tissue boundaries due to low image quality. In particular, in echocardiography, sparse annotations make it challenging to predict landmarks with position stability and temporal consistency. In this paper, we propose a spatio-temporal graph convolutional network tailored for echocardiography landmark detection. We specifically sample landmark labels from the left ventricular endocardium and pre-calculate their correlations to establish structural priors. Our approach involves a graph convolutional neural network that learns the interrelationships among landmarks, significantly enhancing landmark accuracy within ambiguous tissue contexts. Additionally, we integrate gate recurrent units to grasp the temporal consistency of landmarks across consecutive images, augmenting the model's resilience against unlabeled data. Through validation across three echocardiography datasets, our method demonstrates superior accuracy when contrasted with alternative landmark detection models.

PMID:39024972 | DOI:10.1016/j.media.2024.103272

Comput Biol Med. 2024 Jul 17;179:108870. doi: 10.1016/j.compbiomed.2024.108870. Online ahead of print.

ABSTRACT

Accurate segmentation of the liver and tumors from CT volumes is crucial for hepatocellular carcinoma diagnosis and pre-operative resection planning. Despite advances in deep learning-based methods for abdominal CT images, fully-automated segmentation remains challenging due to class imbalance and structural variations, often requiring cascaded approaches that incur significant computational costs. In this paper, we present the Dual-Encoder Double Concatenation Network (DEDC-Net) for simultaneous segmentation of the liver and its tumors. DEDC-Net leverages both residual and skip connections to enhance feature reuse and optimize performance in liver and tumor segmentation tasks. Extensive qualitative and quantitative experiments on the LiTS dataset demonstrate that DEDC-Net outperforms existing state-of-the-art liver segmentation methods. An ablation study was conducted to evaluate different encoder backbones - specifically VGG19 and ResNet - and the impact of incorporating an attention mechanism. Our results indicate that DEDC-Net, without any additional attention gates, achieves a superior mean Dice Score (DS) of 0.898 for liver segmentation. Moreover, integrating residual connections into one encoder yielded the highest DS for tumor segmentation tasks. The robustness of our proposed network was further validated on two additional, unseen CT datasets: IDCARDb-01 and COMET. Our model demonstrated superior lesion segmentation capabilities, particularly on IRCADb-01, achieving a DS of 0.629. The code implementation is publicly available at this website.

PMID:39024904 | DOI:10.1016/j.compbiomed.2024.108870

BMC Med Inform Decis Mak. 2024 Jul 18;24(1):196. doi: 10.1186/s12911-024-02601-4.

ABSTRACT

BACKGROUND: Generalized Joint Hyper-mobility (GJH) can aid in the diagnosis of Ehlers-Danlos Syndrome (EDS), a complex genetic connective tissue disorder with clinical features that can mimic other disease processes. Our study focuses on developing a unique image-based goniometry system, the HybridPoseNet, which utilizes a hybrid deep learning model.

OBJECTIVE: The proposed model is designed to provide the most accurate joint angle measurements in EDS appraisals. Using a hybrid of CNNs and HyperLSTMs in the pose estimation module of HybridPoseNet offers superior generalization and time consistency properties, setting it apart from existing complex libraries.

METHODOLOGY: HybridPoseNet integrates the spatial pattern recognition prowess of MobileNet-V2 with the sequential data processing capability of HyperLSTM units. The system captures the dynamic nature of joint motion by creating a model that learns from individual frames and the sequence of movements. The CNN module of HybridPoseNet was trained on a large and diverse data set before the fine-tuning of video data involving 50 individuals visiting the EDS clinic, focusing on joints that can hyperextend. HyperLSTMs have been incorporated in video frames to avoid any time breakage in joint angle estimation in consecutive frames. The model performance was evaluated using Spearman's coefficient correlation versus manual goniometry measurements, as well as by the human labeling of joint position, the second validation step.

OUTCOME: Preliminary findings demonstrate HybridPoseNet achieving a remarkable correlation with manual Goniometric measurements: thumb (rho = 0.847), elbows (rho = 0.822), knees (rho = 0.839), and fifth fingers (rho = 0.896), indicating that the newest model is considerably better. The model manifested a consistent performance in all joint assessments, hence not requiring selecting a variety of pose-measuring libraries for every joint. The presentation of HybridPoseNet contributes to achieving a combined and normalized approach to reviewing the mobility of joints, which has an overall enhancement of approximately 20% in accuracy compared to the regular pose estimation libraries. This innovation is very valuable to the field of medical diagnostics of connective tissue diseases and a vast improvement to its understanding.

PMID:39026270 | DOI:10.1186/s12911-024-02601-4

BMC Cancer. 2024 Jul 18;24(1):852. doi: 10.1186/s12885-024-12575-1.

ABSTRACT

BACKGROUND: Providing appropriate specialized treatment to the right patient at the right time is considered necessary in cancer management. Targeted therapy tailored to the genetic changes of each breast cancer patient is a desirable feature of precision oncology, which can not only reduce disease progression but also potentially increase patient survival. The use of artificial intelligence alongside precision oncology can help physicians by identifying and selecting more effective treatment factors for patients.

METHOD: A systematic review was conducted using the PubMed, Embase, Scopus, and Web of Science databases in September 2023. We performed the search strategy with keywords, namely: Breast Cancer, Artificial intelligence, and precision Oncology along with their synonyms in the article titles. Descriptive, qualitative, review, and non-English studies were excluded. The quality assessment of the articles and evaluation of bias were determined based on the SJR journal and JBI indices, as well as the PRISMA2020 guideline.

RESULTS: Forty-six studies were selected that focused on personalized breast cancer management using artificial intelligence models. Seventeen studies using various deep learning methods achieved a satisfactory outcome in predicting treatment response and prognosis, contributing to personalized breast cancer management. Two studies utilizing neural networks and clustering provided acceptable indicators for predicting patient survival and categorizing breast tumors. One study employed transfer learning to predict treatment response. Twenty-six studies utilizing machine-learning methods demonstrated that these techniques can improve breast cancer classification, screening, diagnosis, and prognosis. The most frequent modeling techniques used were NB, SVM, RF, XGBoost, and Reinforcement Learning. The average area under the curve (AUC) for the models was 0.91. Moreover, the average values for accuracy, sensitivity, specificity, and precision were reported to be in the range of 90-96% for the models.

CONCLUSION: Artificial intelligence has proven to be effective in assisting physicians and researchers in managing breast cancer treatment by uncovering hidden patterns in complex omics and genetic data. Intelligent processing of omics data through protein and gene pattern classification and the utilization of deep neural patterns has the potential to significantly transform the field of complex disease management.

PMID:39026174 | DOI:10.1186/s12885-024-12575-1

Eur Radiol. 2024 Jul 18. doi: 10.1007/s00330-024-10967-2. Online ahead of print.

ABSTRACT

OBJECTIVES: The aim of this study is to develop a deep-learning model to create synthetic temporal bone computed tomography (CT) images from ultrashort echo-time magnetic resonance imaging (MRI) scans, thereby addressing the intrinsic limitations of MRI in localizing anatomic landmarks in temporal bone CT.

MATERIALS AND METHODS: This retrospective study included patients who underwent temporal MRI and temporal bone CT within one month between April 2020 and March 2023. These patients were randomly divided into training and validation datasets. A CycleGAN model for generating synthetic temporal bone CT images was developed using temporal bone CT and pointwise encoding-time reduction with radial acquisition (PETRA). To assess the model's performance, the pixel count in mastoid air cells was measured. Two neuroradiologists evaluated the successful generation rates of 11 anatomical landmarks.

RESULTS: A total of 102 patients were included in this study (training dataset, n = 54, mean age 58 ± 14, 34 females (63%); validation dataset, n = 48, mean age 61 ± 13, 29 females (60%)). In the pixel count of mastoid air cells, no difference was observed between synthetic and real images (679 ± 342 vs 738 ± 342, p = 0.13). For the six major anatomical sites, the positive generation rates were 97-100%, whereas those of the five major anatomical structures ranged from 24% to 83%.

CONCLUSION: We developed a model to generate synthetic temporal bone CT images using PETRA MRI. This model can provide information regarding the major anatomic sites of the temporal bone using MRI.

CLINICAL RELEVANCE STATEMENT: The proposed algorithm addresses the primary limitations of MRI in localizing anatomic sites within the temporal bone.

KEY POINTS: CT is preferred for imaging the temporal bone, but has limitations in differentiating pathology there. The model achieved a high success rate in generating synthetic images of six anatomic sites. This can overcome the limitations of MRI in visualizing key anatomic sites in the temporal skull.

PMID:39026063 | DOI:10.1007/s00330-024-10967-2

Med Phys. 2024 Jul 18. doi: 10.1002/mp.17308. Online ahead of print.

ABSTRACT

BACKGROUND: The current level of automation in the production of radiotherapy plans for lung cancer patients is relatively low. With the development of artificial intelligence, it has become a reality to use neural networks to predict dose distributions and provide assistance for radiation therapy planning. However, due to the significant individual variability in the distribution of non-small cell lung cancer (NSCLC) planning target volume (PTV) and the complex spatial relationships between the PTV and organs at risk (OARs), there is still a lack of a high-precision dose prediction network tailored to the characteristics of NSCLC.

PURPOSE: To assist in the development of volumetric modulated arc therapy (VMAT) plans for non-small cell lung cancer patients, a deep neural network is proposed to predict high-precision dose distribution.

METHODS: This study has developed a network called MHA-ResUNet, which combines a large-kernel dilated convolution module and multi-head attention (MHA) modules. The network was trained based on 80 VMAT plans of NSCLC patients. CT images, PTV, and OARs were fed into the independent input channel. The dose distribution was taken as the output to train the model. The performance of this network was compared with that of several commonly used networks, and the networks' performance was evaluated based on the voxel-level mean absolute error (MAE) within the PTV and OARs, as well as the error in clinical dose-volume metrics.

RESULTS: The MAE between the predicted dose distribution and the manually planned dose distribution within the PTV is 1.43 Gy, and the D95 error is less than 1 Gy. Compared with the other three commonly used networks, the dose error of the MHA-ResUNet is the smallest in PTV and OARs.

CONCLUSIONS: The proposed MHA-ResUNet network improves the receptive field and filters the shallow features to learn the relative spatial relation between the PTV and the OARs, enabling accurate prediction of dose distributions in NSCLC patients undergoing VMAT radiotherapy.

PMID:39024495 | DOI:10.1002/mp.17308

Dentomaxillofac Radiol. 2024 Jul 18:twae033. doi: 10.1093/dmfr/twae033. Online ahead of print.

ABSTRACT

OBJECTIVES: This study aimed to propose a new method for the automatic diagnosis of anterior disc displacement of the temporomandibular joint (TMJ) using magnetic resonance imaging (MRI) and deep learning. By employing a multistage approach, the factors affecting the final result can be easily identified and improved.

METHODS: This study introduces a multistage automatic diagnostic technique using deep learning. This process involves segmenting the target from MR images, extracting distance parameters, and classifying the diagnosis into three classes. MRI exams of 368 TMJs from 204 patients were evaluated for anterior disc displacement. In the first stage, five algorithms were used for the semantic segmentation of the disc and condyle. In the second stage, 54 distance parameters were extracted from the segments. In the third stage, a rule-based decision model was developed to link the parameters with the expert diagnosis results.

RESULTS: In the first stage, DeepLabV3+ showed the best result (95% Hausdorff distance, Dice coefficient, and sensitivity of 6.47 ± 7.22, 0.84 ± 0.07, and 0.84 ± 0.09, respectively). This study used the original MRI exams as input without preprocessing and showed high segmentation performance compared with that of previous studies. In the third stage, the combination of SegNet and a random forest model yielded an accuracy of 0.89 ± 0.06.

CONCLUSIONS: An algorithm was developed to automatically diagnose TMJ-anterior disc displacement using MRI. Through a multistage approach, this algorithm facilitated the improvement of results and demonstrated high accuracy from more complex inputs. Furthermore, existing radiological knowledge was applied and validated.

PMID:39024472 | DOI:10.1093/dmfr/twae033

Science. 2024 Jul 19;385(6706):276-282. doi: 10.1126/science.adn3780. Epub 2024 Jul 18.

ABSTRACT

We describe an approach for designing high-affinity small molecule-binding proteins poised for downstream sensing. We use deep learning-generated pseudocycles with repeating structural units surrounding central binding pockets with widely varying shapes that depend on the geometry and number of the repeat units. We dock small molecules of interest into the most shape complementary of these pseudocycles, design the interaction surfaces for high binding affinity, and experimentally screen to identify designs with the highest affinity. We obtain binders to four diverse molecules, including the polar and flexible methotrexate and thyroxine. Taking advantage of the modular repeat structure and central binding pockets, we construct chemically induced dimerization systems and low-noise nanopore sensors by splitting designs into domains that reassemble upon ligand addition.

PMID:39024436 | DOI:10.1126/science.adn3780

PLoS Comput Biol. 2024 Jul 18;20(7):e1012283. doi: 10.1371/journal.pcbi.1012283. Online ahead of print.

ABSTRACT

All fields of science depend on mathematical models. Occam's razor refers to the principle that good models should exclude parameters beyond those minimally required to describe the systems they represent. This is because redundancy can lead to incorrect estimates of model parameters from data, and thus inaccurate or ambiguous conclusions. Here, we show how deep learning can be powerfully leveraged to apply Occam's razor to model parameters. Our method, FixFit, uses a feedforward deep neural network with a bottleneck layer to characterize and predict the behavior of a given model from its input parameters. FixFit has three major benefits. First, it provides a metric to quantify the original model's degree of complexity. Second, it allows for the unique fitting of data. Third, it provides an unbiased way to discriminate between experimental hypotheses that add value versus those that do not. In three use cases, we demonstrate the broad applicability of this method across scientific domains. To validate the method using a known system, we apply FixFit to recover known composite parameters for the Kepler orbit model and a dynamic model of blood glucose regulation. In the latter, we demonstrate the ability to fit the latent parameters to real data. To illustrate how the method can be applied to less well-established fields, we use it to identify parameters for a multi-scale brain model and reduce the search space for viable candidate mechanisms.

PMID:39024398 | DOI:10.1371/journal.pcbi.1012283

PLoS One. 2024 Jul 18;19(7):e0307187. doi: 10.1371/journal.pone.0307187. eCollection 2024.

ABSTRACT

In the urban scene segmentation, the "image-to-image translation issue" refers to the fundamental task of transforming input images into meaningful segmentation maps, which essentially involves translating the visual information present in the input image into semantic labels for different classes. When this translation process is inaccurate or incomplete, it can lead to failed segmentation results where the model struggles to correctly classify pixels into the appropriate semantic categories. The study proposed a conditional Generative Adversarial Network (cGAN), for creating high-resolution urban maps from satellite images. The method combines semantic and spatial data using cGAN framework to produce realistic urban scenes while maintaining crucial details. To assess the performance of the proposed method, extensive experiments are performed on benchmark datasets, the ISPRS Potsdam and Vaihingen datasets. Intersection over Union (IoU) and Pixel Accuracy are two quantitative metrics used to evaluate the segmentation accuracy of the produced maps. The proposed method outperforms traditional methods with an IoU of 87% and a Pixel Accuracy of 93%. The experimental findings show that the suggested cGAN-based method performs better than traditional techniques, attaining better segmentation accuracy and generating better urban maps with finely detailed information. The suggested approach provides a framework for resolving the image-to-image translation difficulties in urban scene segmentation, demonstrating the potential of cGANs for producing excellent urban maps from satellite data.

PMID:39024353 | DOI:10.1371/journal.pone.0307187

J Chem Theory Comput. 2024 Jul 18. doi: 10.1021/acs.jctc.4c00594. Online ahead of print.

ABSTRACT

Electron density-based implicit solvation models are a class of techniques for quantifying solvation effects and calculating free energies of solvation without an explicit representation of solvent molecules. Integral to the accuracy of solvation modeling is the proper definition of the solvation shell separating the solute molecule from the solvent environment, allowing for a physical partitioning of the free energies of solvation. Unlike state-of-the-art implicit solvation models for molecular quantum chemistry calculations, e.g., the solvation model based on solute electron density (SMD), solvation models for systems under periodic boundary conditions with plane-wave (PW) basis sets have been limited in their accuracy. Furthermore, a unified implicit solvation model with both homogeneous solution-phase and heterogeneous interfacial structures treated on equal footing is needed. In order to address this challenge, we developed a high-accuracy solvation model for periodic PW calculations that is applicable to molecular, ionic, interfacial, and bulk-phase chemistry. Our model, PW-SMD, is an extension of the SMD molecular solvation model to periodic systems in water. The free energy of solvation is partitioned into the electrostatic and cavity-dispersion-solvent structure (CDS) contributions. The electrostatic contributions of the solvation shell surrounding solute structures are parametrized based on their geometric and physical properties. In addition, the nonelectrostatic contribution to the solvation energy is accounted for by extending the CDS formalism of SMD to incorporate periodic boundary conditions. We validate the accuracy and robustness of our solvation model by comparing predicted solvation free energies against experimental data for molecular and ionic systems, carved-cluster composite energetic models of solvated reaction energies and barriers on surface systems, and deep-learning-accelerated ab initio molecular dynamics (AIMD). Our developed periodic implicit solvation model shows significantly improved accuracy compared to previous work (namely, solvation models in aqueous solution) and can be applied to simulate solvent effects in a wide range of surface and crystalline materials.

PMID:39024317 | DOI:10.1021/acs.jctc.4c00594

PLoS One. 2024 Jul 18;19(7):e0305654. doi: 10.1371/journal.pone.0305654. eCollection 2024.

ABSTRACT

Even though deep learning shows impressive results in several applications, its use on problems with High Dimensions and Low Sample Size, such as diagnosing rare diseases, leads to overfitting. One solution often proposed is feature selection. In deep learning, along with feature selection, network sparsification is also used to improve the results when dealing with high dimensions low sample size data. However, most of the time, they are tackled as separate problems. This paper proposes a new approach that integrates feature selection, based on sparsification, into the training process of a deep neural network. This approach uses a constrained biobjective gradient descent method. It provides a set of Pareto optimal neural networks that make a trade-off between network sparsity and model accuracy. Results on both artificial and real datasets show that using a constrained biobjective gradient descent increases the network sparsity without degrading the classification performances. With the proposed approach, on an artificial dataset, the feature selection score reached 0.97 with a sparsity score of 0.92 with an accuracy of 0.9. For the same accuracy, none of the other methods reached a feature score above 0.20 and sparsity score of 0.35. Finally, statistical tests validate the results obtained on all datasets.

PMID:39024199 | DOI:10.1371/journal.pone.0305654

Br J Haematol. 2024 Jul 18. doi: 10.1111/bjh.19621. Online ahead of print.

ABSTRACT

Palpebral conjunctival hue alteration is used in non-invasive screening for anaemia, whereas it is a qualitative measure. This study constructed machine/deep learning models for predicting haemoglobin values using 150 palpebral conjunctival images taken by a smartphone. The median haemoglobin value was 13.1 g/dL, including 10 patients with <11 g/dL. A segmentation model using U-net was successfully constructed. The segmented images were subjected to non-convolutional neural network (CNN)-based and CNN-based regression models for predicting haemoglobin values. The correlation coefficients between the actual and predicted haemoglobin values were 0.38 and 0.44 in the non-CNN-based and CNN-based models, respectively. The sensitivity and specificity for anaemia detection were 13% and 98% for the non-CNN-based model and 20% and 99% for the CNN-based model. The performance of the CNN-based model did not improve with a mask layer guiding the model's attention towards the conjunctival regions, however, slightly improved with correction by the aspect ratio and exposure time of input images. The gradient-weighted class activation mapping heatmap indicated that the lower half area of the conjunctiva was crucial for haemoglobin value prediction. In conclusion, the CNN-based model had better results than the non-CNN-based model. The prediction accuracy would improve by using more input data with anaemia.

PMID:39024119 | DOI:10.1111/bjh.19621

IEEE Trans Neural Netw Learn Syst. 2024 Jul 18;PP. doi: 10.1109/TNNLS.2024.3420218. Online ahead of print.

ABSTRACT

Neurosymbolic artificial intelligence (AI) is an increasingly active area of research that combines symbolic reasoning methods with deep learning to leverage their complementary benefits. As knowledge graphs (KGs) are becoming a popular way to represent heterogeneous and multirelational data, methods for reasoning on graph structures have attempted to follow this neurosymbolic paradigm. Traditionally, such approaches have utilized either rule-based inference or generated representative numerical embeddings from which patterns could be extracted. However, several recent studies have attempted to bridge this dichotomy to generate models that facilitate interpretability, maintain competitive performance, and integrate expert knowledge. Therefore, we survey methods that perform neurosymbolic reasoning tasks on KGs and propose a novel taxonomy by which we can classify them. Specifically, we propose three major categories: 1) logically informed embedding approaches; 2) embedding approaches with logical constraints; and 3) rule-learning approaches. Alongside the taxonomy, we provide a tabular overview of the approaches and links to their source code, if available, for more direct comparison. Finally, we discuss the unique characteristics and limitations of these methods and then propose several prospective directions toward which this field of research could evolve.

PMID:39024082 | DOI:10.1109/TNNLS.2024.3420218

Dentomaxillofac Radiol. 2024 Jul 18:twae036. doi: 10.1093/dmfr/twae036. Online ahead of print.

ABSTRACT

OBJECTIVE: This study aimed to assess the effectiveness of deep convolutional neural network (CNN) algorithms in detecting and segmentation of overhanging dental restorations in bitewing radiographs.

METHOD AND MATERIALS: A total of 1160 anonymized bitewing radiographs were used to progress the Artificial Intelligence (AI) system) for the detection and segmentation of overhanging restorations. The data was then divided into three groups: 80% for training (930 images, 2399 labels), 10% for validation (115 images, 273 labels), and 10% for testing (115 images, 306 labels). A CNN model known as you only look once (YOLOv5) was trained to detect overhanging restorations in bitewing radiographs. After utilizing the remaining 115 radiographs to evaluate the efficacy of the proposed CNN model, the accuracy, sensitivity, precision, F1 score, and area under the receiver operating characteristic curve (AUC) were computed.

RESULTS: The model demonstrated a precision of 90.9%, a sensitivity of 85.3%, and an F1 score of 88.0%. Furthermore, the model achieved an AUC of 0.859 on the Receiver Operating Characteristic (ROC) curve. The mean average precision (mAP) at an intersection over a union (IoU) threshold of 0.5 was notably high at 0.87.

CONCLUSION: The findings suggest that deep CNN algorithms are highly effective in the detection and diagnosis of overhanging dental restorations in bitewing radiographs. The high levels of precision, sensitivity, and F1 score, along with the significant AUC and mAP values, underscore the potential of these advanced deep learning techniques in revolutionizing dental diagnostic procedures.

PMID:39024043 | DOI:10.1093/dmfr/twae036

JMIR Public Health Surveill. 2024 Jul 18;10:e52353. doi: 10.2196/52353.

ABSTRACT

BACKGROUND: Multimorbidity is a significant public health concern, characterized by the coexistence and interaction of multiple preexisting medical conditions. This complex condition has been associated with an increased risk of COVID-19. Individuals with multimorbidity who contract COVID-19 often face a significant reduction in life expectancy. The postpandemic period has also highlighted an increase in frailty, emphasizing the importance of integrating existing multimorbidity details into epidemiological risk assessments. Managing clinical data that include medical histories presents significant challenges, particularly due to the sparsity of data arising from the rarity of multimorbidity conditions. Also, the complex enumeration of combinatorial multimorbidity features introduces challenges associated with combinatorial explosions.

OBJECTIVE: This study aims to assess the severity of COVID-19 in individuals with multiple medical conditions, considering their demographic characteristics such as age and sex. We propose an evolutionary machine learning model designed to handle sparsity, analyzing preexisting multimorbidity profiles of patients hospitalized with COVID-19 based on their medical history. Our objective is to identify the optimal set of multimorbidity feature combinations strongly associated with COVID-19 severity. We also apply the Apriori algorithm to these evolutionarily derived predictive feature combinations to identify those with high support.

METHODS: We used data from 3 administrative sources in Piedmont, Italy, involving 12,793 individuals aged 45-74 years who tested positive for COVID-19 between February and May 2020. From their 5-year pre-COVID-19 medical histories, we extracted multimorbidity features, including drug prescriptions, disease diagnoses, sex, and age. Focusing on COVID-19 hospitalization, we segmented the data into 4 cohorts based on age and sex. Addressing data imbalance through random resampling, we compared various machine learning algorithms to identify the optimal classification model for our evolutionary approach. Using 5-fold cross-validation, we evaluated each model's performance. Our evolutionary algorithm, utilizing a deep learning classifier, generated prediction-based fitness scores to pinpoint multimorbidity combinations associated with COVID-19 hospitalization risk. Eventually, the Apriori algorithm was applied to identify frequent combinations with high support.

RESULTS: We identified multimorbidity predictors associated with COVID-19 hospitalization, indicating more severe COVID-19 outcomes. Frequently occurring morbidity features in the final evolved combinations were age>53, R03BA (glucocorticoid inhalants), and N03AX (other antiepileptics) in cohort 1; A10BA (biguanide or metformin) and N02BE (anilides) in cohort 2; N02AX (other opioids) and M04AA (preparations inhibiting uric acid production) in cohort 3; and G04CA (Alpha-adrenoreceptor antagonists) in cohort 4.

CONCLUSIONS: When combined with other multimorbidity features, even less prevalent medical conditions show associations with the outcome. This study provides insights beyond COVID-19, demonstrating how repurposed administrative data can be adapted and contribute to enhanced risk assessment for vulnerable populations.

PMID:39024001 | DOI:10.2196/52353