NPJ Digit Med. 2024 Jun 20;7(1):162. doi: 10.1038/s41746-024-01159-9.

ABSTRACT

Middle-ear conditions are common causes of primary care visits, hearing impairment, and inappropriate antibiotic use. Deep learning (DL) may assist clinicians in interpreting otoscopic images. This study included patients over 5 years old from an ambulatory ENT practice in Strasbourg, France, between 2013 and 2020. Digital otoscopic images were obtained using a smartphone-attached otoscope (Smart Scope, Karl Storz, Germany) and labeled by a senior ENT specialist across 11 diagnostic classes (reference standard). An Inception-v2 DL model was trained using 41,664 otoscopic images, and its diagnostic accuracy was evaluated by calculating class-specific estimates of sensitivity and specificity. The model was then incorporated into a smartphone app called i-Nside. The DL model was evaluated on a validation set of 3,962 images and a held-out test set comprising 326 images. On the validation set, all class-specific estimates of sensitivity and specificity exceeded 98%. On the test set, the DL model achieved a sensitivity of 99.0% (95% confidence interval: 94.5-100) and a specificity of 95.2% (91.5-97.6) for the binary classification of normal vs. abnormal images; wax plugs were detected with a sensitivity of 100% (94.6-100) and specificity of 97.7% (95.0-99.1); other class-specific estimates of sensitivity and specificity ranged from 33.3% to 92.3% and 96.0% to 100%, respectively. We present an end-to-end DL-enabled system able to achieve expert-level diagnostic accuracy for identifying normal tympanic aspects and wax plugs within digital otoscopic images. However, the system's performance varied for other middle-ear conditions. Further prospective validation is necessary before wider clinical deployment.

PMID:38902477 | DOI:10.1038/s41746-024-01159-9

Eur Radiol Exp. 2024 Jun 21;8(1):67. doi: 10.1186/s41747-024-00470-0.

ABSTRACT

BACKGROUND: We compared magnetic resonance imaging (MRI) turbo spin-echo images reconstructed using a deep learning technique (TSE-DL) with standard turbo spin-echo (TSE-SD) images of the lumbar spine regarding image quality and detection performance of common degenerative pathologies.

METHODS: This prospective, single-center study included 31 patients (15 males and 16 females; aged 51 ± 16 years (mean ± standard deviation)) who underwent lumbar spine exams with both TSE-SD and TSE-DL acquisitions for degenerative spine diseases. Images were analyzed by two radiologists and assessed for qualitative image quality using a 4-point Likert scale, quantitative signal-to-noise ratio (SNR) of anatomic landmarks, and detection of common pathologies. Paired-sample t, Wilcoxon, and McNemar tests, unweighted/linearly weighted Cohen κ statistics, and intraclass correlation coefficients were used.

RESULTS: Scan time for TSE-DL and TSE-SD protocols was 2:55 and 5:17 min:s, respectively. The overall image quality was either significantly higher for TSE-DL or not significantly different between TSE-SD and TSE-DL. TSE-DL demonstrated higher SNR and subject noise scores than TSE-SD. For pathology detection, the interreader agreement was substantial to almost perfect for TSE-DL, with κ values ranging from 0.61 to 1.00; the interprotocol agreement was almost perfect for both readers, with κ values ranging from 0.84 to 1.00. There was no significant difference in the diagnostic confidence or detection rate of common pathologies between the two sequences (p ≥ 0.081).

CONCLUSIONS: TSE-DL allowed for a 45% reduction in scan time over TSE-SD in lumbar spine MRI without compromising the overall image quality and showed comparable detection performance of common pathologies in the evaluation of degenerative lumbar spine changes.

RELEVANCE STATEMENT: Deep learning-reconstructed lumbar spine MRI protocol enabled a 45% reduction in scan time compared with conventional reconstruction, with comparable image quality and detection performance of common degenerative pathologies.

KEY POINTS: • Lumbar spine MRI with deep learning reconstruction has broad application prospects. • Deep learning reconstruction of lumbar spine MRI saved 45% scan time without compromising overall image quality. • When compared with standard sequences, deep learning reconstruction showed similar detection performance of common degenerative lumbar spine pathologies.

PMID:38902467 | DOI:10.1186/s41747-024-00470-0

Skeletal Radiol. 2024 Jun 20. doi: 10.1007/s00256-024-04684-6. Online ahead of print.

ABSTRACT

This article will provide a perspective review of the most extensively investigated deep learning (DL) applications for musculoskeletal disease detection that have the best potential to translate into routine clinical practice over the next decade. Deep learning methods for detecting fractures, estimating pediatric bone age, calculating bone measurements such as lower extremity alignment and Cobb angle, and grading osteoarthritis on radiographs have been shown to have high diagnostic performance with many of these applications now commercially available for use in clinical practice. Many studies have also documented the feasibility of using DL methods for detecting joint pathology and characterizing bone tumors on magnetic resonance imaging (MRI). However, musculoskeletal disease detection on MRI is difficult as it requires multi-task, multi-class detection of complex abnormalities on multiple image slices with different tissue contrasts. The generalizability of DL methods for musculoskeletal disease detection on MRI is also challenging due to fluctuations in image quality caused by the wide variety of scanners and pulse sequences used in routine MRI protocols. The diagnostic performance of current DL methods for musculoskeletal disease detection must be further evaluated in well-designed prospective studies using large image datasets acquired at different institutions with different imaging parameters and imaging hardware before they can be fully implemented in clinical practice. Future studies must also investigate the true clinical benefits of current DL methods and determine whether they could enhance quality, reduce error rates, improve workflow, and decrease radiologist fatigue and burnout with all of this weighed against the costs.

PMID:38902420 | DOI:10.1007/s00256-024-04684-6

Sci Rep. 2024 Jun 20;14(1):14196. doi: 10.1038/s41598-024-61452-1.

ABSTRACT

Alzheimer disease (AD) is among the most chronic neurodegenerative diseases that threaten global public health. The prevalence of Alzheimer disease and consequently the increased risk of spread all over the world pose a vital threat to human safekeeping. Early diagnosis of AD is a suitable action for timely intervention and medication, which may increase the prognosis and quality of life for affected individuals. Quantum computing provides a more efficient model for different disease classification tasks than classical machine learning approaches. The full potential of quantum computing is not applied to Alzheimer's disease classification tasks as expected. In this study, we proposed an ensemble deep learning model based on quantum machine learning classifiers to classify Alzheimer's disease. The Alzheimer's disease Neuroimaging Initiative I and Alzheimer's disease Neuroimaging Initiative II datasets are merged for the AD disease classification. We combined important features extracted based on the customized version of VGG16 and ResNet50 models from the merged images then feed these features to the Quantum Machine Learning classifier to classify them as non-demented, mild demented, moderate demented, and very mild demented. We evaluate the performance of our model by using six metrics; accuracy, the area under the curve, F1-score, precision, and recall. The result validates that the proposed model outperforms several state-of-the-art methods for detecting Alzheimer's disease by registering an accuracy of 99.89 and 98.37 F1-score.

PMID:38902368 | DOI:10.1038/s41598-024-61452-1

Sci Rep. 2024 Jun 20;14(1):14264. doi: 10.1038/s41598-024-65040-1.

ABSTRACT

The traffic flow prediction is the key to alleviate traffic congestion, yet very challenging due to the complex influence factors. Currently, the most of deep learning models are designed to dig out the intricate dependency in continuous standardized sequences, which are dependent to high requirements for data continuity and regularized distribution. However, the data discontinuity and irregular distribution are inevitable in the real-world practical application, then we need find a way to utilize the powerful effect of the multi-feature fusion rather than continuous relation in standardized sequences. To this end, we conduct the prediction based on the multiple traffic features reflecting the complex influence factors. Firstly, we propose the ATFEM, an adaptive traffic features extraction mechanism, which can select important influence factors to construct joint temporal features matrix and global spatial features matrix according to the traffic condition. In this way, the feature's representation ability can be improved. Secondly, we propose the MFSTN, a multi-feature spatial-temporal fusion network, which include the temporal transformer encoder and graph attention network to obtain the latent representation of spatial-temporal features. Especially, we design the scaled spatial-temporal fusion module, which can automatically learn optimal fusion weights, further adapt to inconsistent spatial-temporal dimensions. Finally, the multi-layer perceptron gets the mapping function between these comprehensive features and traffic flow. This method helps to improve the interpretability of the prediction. Experimental results show that the proposed model outperforms a variety of baselines, and it can accurately predict the traffic flow when the data missing rate is high.

PMID:38902350 | DOI:10.1038/s41598-024-65040-1

SLAS Technol. 2024 Jun 18:100161. doi: 10.1016/j.slast.2024.100161. Online ahead of print.

ABSTRACT

Most classification models for Alzheimer's Diagnosis (AD) do not have specific strategies for individual input samples, leading to the problem of easily overlooking personalized differences between samples. This research introduces a customized dynamically ensemble convolution neural network (PDECNN), which is able to build a specific integration strategy based on the distinctiveness of the sample. In this paper, we propose a personalized dynamic ensemble alzheimer's Diagnosis classification model. This model will dynamically modify the deteriorated brain areas of interest depending on various samples since it can adjust to variations in the degeneration of sample brain areas. In clinical problems, the PDECNN model has additional diagnostic importance since it can identify sample-specific degraded brain areas based on input samples. This model considers the variability of brain region degeneration levels between input samples, evaluates the degree of degeneration of specific brain regions using an attention mechanism, and selects and integrates brain region features based on the degree of degeneration. Furthermore, by redesigning the classification accuracy performance, we respectively improve it by 4%, 11%, and 8%. Moreover, the degraded brain regions identified by the model show high consistency with the clinical manifestations of AD.

PMID:38901762 | DOI:10.1016/j.slast.2024.100161

Can J Ophthalmol. 2024 Jun 17:S0008-4182(24)00157-1. doi: 10.1016/j.jcjo.2024.05.014. Online ahead of print.

ABSTRACT

OBJECTIVE: To compare the visibility and accessibility of the outer retina in neovascular age-related macular degeneration (nAMD) between 2 OCT devices.

METHODS: In this prospective, cross-sectional exploratory study, differences in thickness and loss of individual outer retinal layers in eyes with nAMD and in age-matched healthy eyes between a next-level High-Res OCT device and the conventional SPECTRALIS OCT (both Heidelberg Engineering GmbH, Heidelberg, Germany) were analyzed. Eyes with nAMD and at least 250 nL of retinal fluid, quantified by an approved deep-learning algorithm (Fluid Monitor, RetInSight, Vienna, Austria), fulfilled the inclusion criteria. The outer retinal layers were segmented using automated layer segmentation and were corrected manually. Layer loss and thickness were compared between both devices using a linear mixed-effects model and a paired t test.

RESULTS: Nineteen eyes of 17 patients with active nAMD and 17 healthy eyes were included. For nAMD eyes, the thickness of the retinal pigment epithelium (RPE) differed significantly between the devices (25.42 μm [95% CI, 14.24-36.61] and 27.31 μm [95% CI, 16.12-38.50] for high-resolution OCT and conventional OCT, respectively; p = 0.033). Furthermore, a significant difference was found in the mean relative external limiting membrane loss (p = 0.021). However, the thickness of photoreceptors, RPE integrity loss, and photoreceptor integrity loss did not differ significantly between devices in the central 3 mm. In healthy eyes, a significant difference in both RPE and photoreceptor thickness between devices was shown (p < 0.001).

CONCLUSION: Central RPE thickness was significantly thinner on high-resolution OCT compared with conventional OCT images explained by superior optical separation of the RPE and Bruch's membrane.

PMID:38901467 | DOI:10.1016/j.jcjo.2024.05.014

Biomed Phys Eng Express. 2024 Jun 20. doi: 10.1088/2057-1976/ad5a14. Online ahead of print.

ABSTRACT

&#xD;&#xD;&#xD;&#xD;&#xD;&#xD;Presently, close to two million patients globally succumb to gastrointestinal reflux diseases (GERD). Video endoscopy represents cutting-edge technology in medical imaging, facilitating the diagnosis of various gastrointestinal ailments including stomach ulcers, bleeding, and polyps. However, the abundance of images produced by medical video endoscopy necessitates significant time for doctors to analyze them thoroughly, posing a challenge for manual diagnosis. This challenge has spurred research into computer-aided techniques aimed at diagnosing the plethora of generated images swiftly and accurately. The novelty of the proposed methodology lies in the development of a system tailored for the diagnosis of gastrointestinal diseases. The proposed work used an object detection method called Yolov5 for identifying abnormal region of interest and Deep LabV3+ for segmentation of abnormal regions in GERD. Further, the features are extracted from the segmented image and given as an input to the seven different machine learning classifiers and custom deep neural network model for multi-stage classification of GERD. The DeepLabV3+ attains an excellent segmentation accuracy of 95.2% and an F1 score of 93.3%. The custom dense neural network obtained a classification accuracy of 90.5%. Among the seven different machine learning classifiers, support vector machine (SVM) outperformed with classification accuracy of 87% compared to all other classifiers. Thus, the combination of object detection, deep learning-based segmentation and machine learning classification enables the timely identification and surveillance of problems associated with GERD for healthcare providers.&#xD;&#xD.

PMID:38901416 | DOI:10.1088/2057-1976/ad5a14

Bioorg Chem. 2024 Jun 10;150:107553. doi: 10.1016/j.bioorg.2024.107553. Online ahead of print.

ABSTRACT

The overexpression of FGFR1 is thought to significantly contribute to the progression of triple-negative breast cancer (TNBC), impacting aspects such as tumorigenesis, growth, metastasis, and drug resistance. Consequently, the pursuit of effective inhibitors for FGFR1 is a key area of research interest. In response to this need, our study developed a hybrid virtual screening method. Utilizing KarmaDock, an innovative algorithm that blends deep learning with molecular docking, alongside Schrödinger's Residue Scanning. This strategy led us to identify compound 6, which demonstrated promising FGFR1 inhibitory activity, evidenced by an IC50 value of approximately 0.24 nM in the HTRF bioassay. Further evaluation revealed that this compound also inhibits the FGFR1 V561M variant with an IC50 value around 1.24 nM. Our subsequent investigations demonstrate that Compound 6 robustly suppresses the migration and invasion capacities of TNBC cell lines, through the downregulation of p-FGFR1 and modulation of EMT markers, highlighting its promise as a potent anti-metastatic therapeutic agent. Additionally, our use of molecular dynamics simulations provided a deeper understanding of the compound's specific binding interactions with FGFR1.

PMID:38901279 | DOI:10.1016/j.bioorg.2024.107553

Int J Med Inform. 2024 Jun 13;189:105523. doi: 10.1016/j.ijmedinf.2024.105523. Online ahead of print.

ABSTRACT

BACKGROUND: The surge in emergency head CT imaging and artificial intelligence (AI) advancements, especially deep learning (DL) and convolutional neural networks (CNN), have accelerated the development of computer-aided diagnosis (CADx) for emergency imaging. External validation assesses model generalizability, providing preliminary evidence of clinical potential.

OBJECTIVES: This study systematically reviews externally validated CNN-CADx models for emergency head CT scans, critically appraises diagnostic test accuracy (DTA), and assesses adherence to reporting guidelines.

METHODS: Studies comparing CNN-CADx model performance to reference standard were eligible. The review was registered in PROSPERO (CRD42023411641) and conducted on Medline, Embase, EBM-Reviews and Web of Science following PRISMA-DTA guideline. DTA reporting were systematically extracted and appraised using standardised checklists (STARD, CHARMS, CLAIM, TRIPOD, PROBAST, QUADAS-2).

RESULTS: Six of 5636 identified studies were eligible. The common target condition was intracranial haemorrhage (ICH), and intended workflow roles auxiliary to experts. Due to methodological and clinical between-study variation, meta-analysis was inappropriate. The scan-level sensitivity exceeded 90 % in 5/6 studies, while specificities ranged from 58,0-97,7 %. The SROC 95 % predictive region was markedly broader than the confidence region, ranging above 50 % sensitivity and 20 % specificity. All studies had unclear or high risk of bias and concern for applicability (QUADAS-2, PROBAST), and reporting adherence was below 50 % in 20 of 32 TRIPOD items.

CONCLUSION: 0.01 % of identified studies met the eligibility criteria. The evidence on the DTA of CNN-CADx models for emergency head CT scans remains limited in the scope of this review, as the reviewed studies were scarce, inapt for meta-analysis and undermined by inadequate methodological conduct and reporting. Properly conducted, external validation remains preliminary for evaluating the clinical potential of AI-CADx models, but prospective and pragmatic clinical validation in comparative trials remains most crucial. In conclusion, future AI-CADx research processes should be methodologically standardized and reported in a clinically meaningful way to avoid research waste.

PMID:38901270 | DOI:10.1016/j.ijmedinf.2024.105523

Comput Biol Med. 2024 Jun 19;178:108721. doi: 10.1016/j.compbiomed.2024.108721. Online ahead of print.

ABSTRACT

Since the 2000s, digitalization has been a crucial transformation in our lives. Nevertheless, digitalization brings a bulk of unstructured textual data to be processed, including articles, clinical records, web pages, and shared social media posts. As a critical analysis, the classification task classifies the given textual entities into correct categories. Categorizing documents from different domains is straightforward since the instances are unlikely to contain similar contexts. However, document classification in a single domain is more complicated due to sharing the same context. Thus, we aim to classify medical articles about four common cancer types (Leukemia, Non-Hodgkin Lymphoma, Bladder Cancer, and Thyroid Cancer) by constructing machine learning and deep learning models. We used 383,914 medical articles about four common cancer types collected by the PubMed API. To build classification models, we split the dataset into 70% as training, 20% as testing, and 10% as validation. We built widely used machine-learning (Logistic Regression, XGBoost, CatBoost, and Random Forest Classifiers) and modern deep-learning (convolutional neural networks - CNN, long short-term memory - LSTM, and gated recurrent unit - GRU) models. We computed the average classification performances (precision, recall, F-score) to evaluate the models over ten distinct dataset splits. The best-performing deep learning model(s) yielded a superior F1 score of 98%. However, traditional machine learning models also achieved reasonably high F1 scores, 95% for the worst-performing case. Ultimately, we constructed multiple models to classify articles, which compose a hard-to-classify dataset in the medical domain.

PMID:38901188 | DOI:10.1016/j.compbiomed.2024.108721

Ultrasonics. 2024 Jun 10;142:107358. doi: 10.1016/j.ultras.2024.107358. Online ahead of print.

ABSTRACT

Stiffness measurement using shear wave propagation velocity has been the most common non-invasive method for liver fibrosis assessment. The velocity is captured through a trace recorded by transient ultrasonographic elastography, with the slope indicating the velocity of the wave. However, due to various factors such as noise and shear wave attenuation, detecting shear wave trajectory on wave propagation maps is a challenging task. In this work, we made the first attempt to use deep learning methods for shear wave trajectory detection on wave propagation maps. Specifically, we adopted five deep learning models in this task and evaluated them by using a well-acknowledged metric based on EA-Angular-Score (EAA) and task-specific metric based on Young s-Score (Ys) in the line-detection field. Furthermore, we proposed an end-to-end framework based on a Transformer and Hough transform, named Transformer-enhanced Hough Transform (TEHT). It took a wave propagation map as input image and directly output the slope of the shear wave trajectory. The framework extracts multi-scale local features from wave propagation maps, employs a deformable attention mechanism for feature fusion, identifies the target line using the Hough transform's voting mechanism, and calculates the contribution of each scale through channel attention. Wave propagation maps from 68 patients were utilized in this study, with manual annotation performed by a rater who was trained as a radiologist, serving as the reference value. The evaluation revealed that the SLNet model exhibited F-measure of EA and Ys values as 40.33 % and 40.72 %, respectively, while the TEHT model showed F-measure of EA and Ys values as 80.96 % and 98.00 %, respectively. TEHT yielded significantly better performance than other deep learning models. Moreover, TEHT demonstrated strong concordance with the gold standard, yielding R2 values of 0.967 and 0.968 for velocity and liver stiffness, respectively. The present study therefore suggests the application of the TEHT model for assessing liver fibrosis owing to its superiority among the five deep learning models.

PMID:38901149 | DOI:10.1016/j.ultras.2024.107358

Neural Netw. 2024 Jun 4;178:106420. doi: 10.1016/j.neunet.2024.106420. Online ahead of print.

ABSTRACT

We study the representation capacity of deep hyperbolic neural networks (HNNs) with a ReLU activation function. We establish the first proof that HNNs can ɛ-isometrically embed any finite weighted tree into a hyperbolic space of dimension d at least equal to 2 with prescribed sectional curvature κ<0, for any ɛ>1 (where ɛ=1 being optimal). We establish rigorous upper bounds for the network complexity on an HNN implementing the embedding. We find that the network complexity of HNN implementing the graph representation is independent of the representation fidelity/distortion. We contrast this result against our lower bounds on distortion which any ReLU multi-layer perceptron (MLP) must exert when embedding a tree with L>2d leaves into a d-dimensional Euclidean space, which we show at least Ω(L1/d); independently of the depth, width, and (possibly discontinuous) activation function defining the MLP.

PMID:38901097 | DOI:10.1016/j.neunet.2024.106420

Neural Netw. 2024 Jun 3;178:106429. doi: 10.1016/j.neunet.2024.106429. Online ahead of print.

ABSTRACT

Although recent studies on blind single image super-resolution (SISR) have achieved significant success, most of them typically require supervised training on synthetic low resolution (LR)-high resolution (HR) paired images. This leads to re-training necessity for different degradations and restricted applications in real-world scenarios with unfavorable inputs. In this paper, we propose an unsupervised blind SISR method with input underlying different degradations, named different degradations blind super-resolution (DDSR). It formulates a Gaussian modeling on blur degradation and employs a meta-learning framework for solving different image degradations. Specifically, a neural network-based kernel generator is optimized by learning from random kernel samples, referred to as random kernel learning. This operation provides effective initialization for blur degradation optimization. At the same time, a meta-learning framework is proposed to resolve multiple degradation modelings on the basis of alternative optimization between blur degradation and image restoration, respectively. Differing from the pre-trained deep-learning methods, the proposed DDSR is implemented in a plug-and-play manner, and is capable of restoring HR image from unfavorable LR input with degradations such as partial coverage, noise addition, and darkening. Extensive simulations illustrate the superior performance of the proposed DDSR approach compared to the state-of-the-arts on public datasets with comparable memory load and time consumption, yet exhibiting better application flexibility and convenience, and significantly better generalization ability towards multiple degradations. Our code is available at https://github.com/XYLGroup/DDSR.

PMID:38901090 | DOI:10.1016/j.neunet.2024.106429

Biomed Eng Online. 2024 Jun 20;23(1):57. doi: 10.1186/s12938-024-01252-w.

ABSTRACT

OBJECTIVE: Our objective was to create a machine learning architecture capable of identifying obstructive sleep apnea (OSA) patterns in single-lead electrocardiography (ECG) signals, exhibiting exceptional performance when utilized in clinical data sets.

METHODS: We conducted our research using a data set consisting of 1656 patients, representing a diverse demographic, from the sleep center of China Medical University Hospital. To detect apnea ECG segments and extract apnea features, we utilized the EfficientNet and some of its layers, respectively. Furthermore, we compared various training and data preprocessing techniques to enhance the model's prediction, such as setting class and sample weights or employing overlapping and regular slicing. Finally, we tested our approach against other literature on the Apnea-ECG database.

RESULTS: Our research found that the EfficientNet model achieved the best apnea segment detection using overlapping slicing and sample-weight settings, with an AUC of 0.917 and an accuracy of 0.855. For patient screening with AHI > 30, we combined the trained model with XGBoost, leading to an AUC of 0.975 and an accuracy of 0.928. Additional tests using PhysioNet data showed that our model is comparable in performance to existing models regarding its ability to screen OSA levels.

CONCLUSIONS: Our suggested architecture, coupled with training and preprocessing techniques, showed admirable performance with a diverse demographic dataset, bringing us closer to practical implementation in OSA diagnosis. Trial registration The data for this study were collected retrospectively from the China Medical University Hospital in Taiwan with approval from the institutional review board CMUH109-REC3-018.

PMID:38902671 | DOI:10.1186/s12938-024-01252-w

JCO Clin Cancer Inform. 2024 Jun;8:e2300184. doi: 10.1200/CCI.23.00184.

ABSTRACT

PURPOSE: Prostate cancer (PCa) represents a highly heterogeneous disease that requires tools to assess oncologic risk and guide patient management and treatment planning. Current models are based on various clinical and pathologic parameters including Gleason grading, which suffers from a high interobserver variability. In this study, we determine whether objective machine learning (ML)-driven histopathology image analysis would aid us in better risk stratification of PCa.

MATERIALS AND METHODS: We propose a deep learning, histopathology image-based risk stratification model that combines clinicopathologic data along with hematoxylin and eosin- and Ki-67-stained histopathology images. We train and test our model, using a five-fold cross-validation strategy, on a data set from 502 treatment-naïve PCa patients who underwent radical prostatectomy (RP) between 2000 and 2012.

RESULTS: We used the concordance index as a measure to evaluate the performance of various risk stratification models. Our risk stratification model on the basis of convolutional neural networks demonstrated superior performance compared with Gleason grading and the Cancer of the Prostate Risk Assessment Post-Surgical risk stratification models. Using our model, 3.9% of the low-risk patients were correctly reclassified to be high-risk and 21.3% of the high-risk patients were correctly reclassified as low-risk.

CONCLUSION: These findings highlight the importance of ML as an objective tool for histopathology image assessment and patient risk stratification. With further validation on large cohorts, the digital pathology risk classification we propose may be helpful in guiding administration of adjuvant therapy including radiotherapy after RP.

PMID:38900978 | DOI:10.1200/CCI.23.00184

PLoS Pathog. 2024 Jun 20;20(6):e1012300. doi: 10.1371/journal.ppat.1012300. Online ahead of print.

ABSTRACT

The AAA-type ATPase VPS4 is recruited by proteins of the endosomal sorting complex required for transport III (ESCRT-III) to catalyse membrane constriction and membrane fission. VPS4A accumulates at the cytoplasmic viral assembly complex (cVAC) of cells infected with human cytomegalovirus (HCMV), the site where nascent virus particles obtain their membrane envelope. Here we show that VPS4A is recruited to the cVAC via interaction with pUL71. Sequence analysis, deep-learning structure prediction, molecular dynamics and mutagenic analysis identify a short peptide motif in the C-terminal region of pUL71 that is necessary and sufficient for the interaction with VPS4A. This motif is predicted to bind the same groove of the N-terminal VPS4A Microtubule-Interacting and Trafficking (MIT) domain as the Type 2 MIT-Interacting Motif (MIM2) of cellular ESCRT-III components, and this viral MIM2-like motif (vMIM2) is conserved across β-herpesvirus pUL71 homologues. However, recruitment of VPS4A by pUL71 is dispensable for HCMV morphogenesis or replication and the function of the conserved vMIM2 during infection remains enigmatic. VPS4-recruitment via a vMIM2 represents a previously unknown mechanism of molecular mimicry in viruses, extending previous observations that herpesviruses encode proteins with structural and functional homology to cellular ESCRT-III components.

PMID:38900818 | DOI:10.1371/journal.ppat.1012300

PLoS Comput Biol. 2024 Jun 20;20(6):e1012231. doi: 10.1371/journal.pcbi.1012231. Online ahead of print.

ABSTRACT

Computational fluid dynamics (CFD) can be used for non-invasive evaluation of hemodynamics. However, its routine use is limited by labor-intensive manual segmentation, CFD mesh creation, and time-consuming simulation. This study aims to train a deep learning model to both generate patient-specific volume-meshes of the pulmonary artery from 3D cardiac MRI data and directly estimate CFD flow fields. This proof-of-concept study used 135 3D cardiac MRIs from both a public and private dataset. The pulmonary arteries in the MRIs were manually segmented and converted into volume-meshes. CFD simulations were performed on ground truth meshes and interpolated onto point-point correspondent meshes to create the ground truth dataset. The dataset was split 110/10/15 for training, validation, and testing. Image2Flow, a hybrid image and graph convolutional neural network, was trained to transform a pulmonary artery template to patient-specific anatomy and CFD values, taking a specific inlet velocity as an additional input. Image2Flow was evaluated in terms of segmentation, and the accuracy of predicted CFD was assessed using node-wise comparisons. In addition, the ability of Image2Flow to respond to increasing inlet velocities was also evaluated. Image2Flow achieved excellent segmentation accuracy with a median Dice score of 0.91 (IQR: 0.86-0.92). The median node-wise normalized absolute error for pressure and velocity magnitude was 11.75% (IQR: 9.60-15.30%) and 9.90% (IQR: 8.47-11.90), respectively. Image2Flow also showed an expected response to increased inlet velocities with increasing pressure and velocity values. This proof-of-concept study has shown that it is possible to simultaneously perform patient-specific volume-mesh based segmentation and pressure and flow field estimation using Image2Flow. Image2Flow completes segmentation and CFD in ~330ms, which is ~5000 times faster than manual methods, making it more feasible in a clinical environment.

PMID:38900817 | DOI:10.1371/journal.pcbi.1012231

STAR Protoc. 2024 Jun 19;5(3):103134. doi: 10.1016/j.xpro.2024.103134. Online ahead of print.

ABSTRACT

Fundus fluorescein angiography (FFA) examinations are widely used in the evaluation of fundus disease conditions to facilitate further treatment suggestions. Here, we present a protocol for performing deep learning-based FFA image analytics with classification and segmentation tasks. We describe steps for data preparation, model implementation, statistical analysis, and heatmap visualization. The protocol is applicable in Python using customized data and can achieve the whole process from diagnosis to treatment suggestion of ischemic retinal diseases. For complete details on the use and execution of this protocol, please refer to Zhao et al.1.

PMID:38900632 | DOI:10.1016/j.xpro.2024.103134

IEEE J Biomed Health Inform. 2024 Jun 20;PP. doi: 10.1109/JBHI.2024.3417229. Online ahead of print.

ABSTRACT

Microscopic cell detection is a challenging task due to significant inter-cell occlusions in dense clusters and diverse cell morphologies. This paper introduces a novel framework designed to enhance automated cell detection. The proposed approach integrates a deep learning model that produces an inverse distance transform-based detection map from the given image, accompanied by a secondary network designed to regress a cell density map from the same input. The inverse distance transform-based map effectively highlights each cell instance in the densely populated areas, while the density map accurately estimates the total cell count in the image. Then, a custom counting-aided cell center extraction strategy leverages the cell count obtained by integrating over the density map to refine the detection process, significantly reducing false responses and thereby boosting overall accuracy. The proposed framework demonstrated superior performance with F-scores of 96.93%, 91.21%, and 92.00% on the VGG, MBM, and ADI datasets, respectively, surpassing existing state-of-the-art methods. It also achieved the lowest distance error, further validating the effectiveness of the proposed approach. These results demonstrate significant potential for automated cell analysis in biomedical applications.

PMID:38900626 | DOI:10.1109/JBHI.2024.3417229