[go: up one dir, main page]
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.5
Journals
Applied Sciences
Special Issues
Application of Artificial Intelligence in Biomedical Informatics
applsci-logo
Submit to Special Issue Submit Abstract to Special Issue Review for Applied Sciences Propose a Special Issue

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Application of Artificial Intelligence in Biomedical Informatics

A special issue of Applied Sciences (ISSN 2076-3417). This special issue belongs to the section "Computing and Artificial Intelligence".

Deadline for manuscript submissions: 10 July 2025 | Viewed by 2935

Special Issue Editor

Dr. Juan Martinez-Romo

E-Mail Website
Guest Editor
Computer Languages and Systems Department, National University of Distance Education, 28006 Madrid, Spain
Interests: machine learning; context-aware inference models; telemedicine applications
Special Issues, Collections and Topics in MDPI journals

Special Issue Information

Dear Colleagues,

Welcome to this Special Issue on the "Application of Artificial Intelligence in Biomedical Informatics". In recent years, the intersection of artificial intelligence (AI) and biomedical informatics has revolutionized healthcare by offering innovative solutions to complex problems. This Special Issue aims to explore the latest advancements, challenges, and opportunities in leveraging AI techniques to enhance biomedical informatics research and applications. From disease diagnosis and personalized medicine to healthcare management and drug discovery, AI is reshaping the landscape of biomedical informatics, promising improved patient outcomes and more efficient healthcare delivery. Join us as we delve into the cutting-edge research and developments driving this transformative field forward.

  • Biomedical informatics;
  • Knowledge-based systems;
  • AI-based clinical decision making;
  • Text mining and health informatics;
  • Natural language processing (NLP) in healthcare;
  • Multilingual clinical NLP;
  • Deep learning applied to healthcare.

Dr. Juan Martinez-Romo
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Applied Sciences is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • biomedical informatics
  • decision making
  • healthcare

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue policies can be found here.

Published Papers (4 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

21 pages, 6202 KiB  
Article
Automatic Meniscus Segmentation Using YOLO-Based Deep Learning Models with Ensemble Methods in Knee MRI Images
by Mehmet Ali Şimşek, Ahmet Sertbaş, Hadi Sasani and Yaşar Mahsut Dinçel
Appl. Sci. 2025, 15(5), 2752; https://doi.org/10.3390/app15052752 - 4 Mar 2025
Viewed by 231
Abstract
The meniscus is a C-shaped connective tissue with a cartilage-like structure in the knee joint. This study proposes an innovative method based on You Only Look Once (YOLO) series models and ensemble methods for meniscus segmentation from knee magnetic resonance imaging (MRI) images [...] Read more.
The meniscus is a C-shaped connective tissue with a cartilage-like structure in the knee joint. This study proposes an innovative method based on You Only Look Once (YOLO) series models and ensemble methods for meniscus segmentation from knee magnetic resonance imaging (MRI) images to improve segmentation performance and evaluate generalization capability. In this study, five different segmentation models were trained, and masks were created from the YOLO series. These masks are combined with pixel-based voting, weighted multiple voting, and dynamic weighted multiple voting optimized by grid search. Tests were conducted on internal and external sets and various metrics. The dynamic weighted multiple voting method optimized with grid search performed the best on both the test set (DSC: 0.8976 ± 0.0071, PPV: 0.8561 ± 0.0121, Sensitivity: 0.9467 ± 0.0077) and the external set (DSC: 0.9004 ± 0.0064, PPV: 0.8876 ± 0.0134, Sensitivity: 0.9200 ± 0.0119). The proposed ensemble methods offer high accuracy, reliability, and generalization capability for meniscus segmentation. Full article
(This article belongs to the Special Issue Application of Artificial Intelligence in Biomedical Informatics)
►▼ Show Figures

Figure 1

Figure 1
<p>General workflow of the proposed segmentation approach.</p> Full article ">Figure 2
<p>The process of creating the internal dataset.</p> Full article ">Figure 3
<p>ROI areas for meniscal segmentation in the sagittal section. (<b>a</b>) An image with meniscal injury in the posterior and anterior horn of the medial meniscus in the test set. (<b>b</b>) Representation of healthy lateral meniscus in the external set.</p> Full article ">Figure 4
<p>Curves showing precision, recall, and F1 score values of the models over 100 epochs. (<b>a</b>) P value curve. (<b>b</b>) R value curve. (<b>c</b>) F1 score value curve.</p> Full article ">Figure 5
<p>Qualitative visual analysis results were obtained by the models in the test set and external set. The image in the test set is the medial meniscus with meniscal damage, the image in the external set is the lateral meniscus with healthy meniscus.</p> Full article ">Figure 6
<p>Masks of the meniscus region predicted by Model 1.</p> Full article ">Figure 7
<p>An example of knee MRI images that represent the sizes of meniscal ROI areas. (<b>a</b>) Medial meniscus with meniscal tear in the internal set. (<b>b</b>) Healthy lateral meniscus in the external set.</p> Full article ">Figure 8
<p>Comparison of segmentation masks obtained by the proposed method (Method 3) and Model 1 with ground truth. (<b>a</b>,<b>b</b>) Internal test images are healthy and torn meniscus samples, respectively. (<b>c</b>,<b>d</b>) External test images are healthy and torn meniscus samples, respectively.</p> Full article ">
20 pages, 6725 KiB  
Article
Novel Deep Learning Method in Hip Osteoarthritis Investigation Before and After Total Hip Arthroplasty
by Roel Pantonial and Milan Simic
Appl. Sci. 2025, 15(2), 872; https://doi.org/10.3390/app15020872 - 17 Jan 2025
Viewed by 570
Abstract
The application of gait analysis on patients with Hip Osteoarthritis (HOA) before and after Total Hip Arthroplasty (THA) surgery can provide accurate diagnostics, reliable treatment decision making, and proper rehabilitation efforts. Acquired kinematic trajectories provide discriminating features that can be used to determine [...] Read more.
The application of gait analysis on patients with Hip Osteoarthritis (HOA) before and after Total Hip Arthroplasty (THA) surgery can provide accurate diagnostics, reliable treatment decision making, and proper rehabilitation efforts. Acquired kinematic trajectories provide discriminating features that can be used to determine the gait patterns of healthy subjects and the effects of surgical operation. However, there is still a lack of consensus on the best discriminating kinematics to achieve this. Our investigation aims to utilize Deep Learning (DL) methodologies and improve classification results for the kinematic parameters of healthy, HOA, and 6 months post-THA gait cycles. Kinematic angles from the lower limb are used directly as one-dimensional inputs into a DL model. Based on the human gait cycle’s features, a hybrid Long Short-Term Memory–Convolutional Neural Network (HLSTM-CNN) is designed for the classification of healthy/HOA/THA gaits. It was found, from the results, that the sagittal angles of hip and knee, and front angles of FPA and knee, provide the most discriminating results with accuracy above 94% between healthy and HOA gaits. Interestingly, when using the sagittal angles of hip and knee to analyze the THA gaits, common subjects have the same results on the misclassifications. This crucial information provides a glimpse in the determination for the success or failure of THA. Full article
(This article belongs to the Special Issue Application of Artificial Intelligence in Biomedical Informatics)
►▼ Show Figures

Figure 1

Figure 1
<p>Image convolution process.</p> Full article ">Figure 2
<p>Typical CNN model.</p> Full article ">Figure 3
<p>LSTM architecture.</p> Full article ">Figure 4
<p>Proposed HLSTM-CNN model.</p> Full article ">Figure 5
<p>Proposed approach.</p> Full article ">Figure 6
<p>Data description: (<b>a</b>) lower limb markers and (<b>b</b>) body plane and angles.</p> Full article ">Figure 7
<p>Aggregated joint angles of the affected limb.</p> Full article ">Figure 8
<p>Gaussian noise addition.</p> Full article ">Figure 9
<p>Data pre-processing flowchart.</p> Full article ">Figure 10
<p>Comparison from the literature: SVM [<a href="#B23-applsci-15-00872" class="html-bibr">23</a>], and CNN (transfer learning) [<a href="#B24-applsci-15-00872" class="html-bibr">24</a>].</p> Full article ">Figure 11
<p>Confusion matrix of the hip sagittal angle.</p> Full article ">Figure 12
<p>Confusion matrix of the knee sagittal angle.</p> Full article ">Figure 13
<p>Confusion matrix of the knee front angle.</p> Full article ">Figure 14
<p>Confusion matrix of the FPA front angle.</p> Full article ">Figure 15
<p>Healthy prediction of subjects after THA.</p> Full article ">Figure 16
<p>HOA prediction of subjects after THA.</p> Full article ">
26 pages, 4034 KiB  
Article
Semi-Supervised Deep Subspace Embedding for Binary Classification of Sella Turcica
by Kaushlesh Singh Shakya, Azadeh Alavi, Julie Porteous, Priti Khatri, Amit Laddi, Manojkumar Jaiswal and Vinay Kumar
Appl. Sci. 2024, 14(23), 11154; https://doi.org/10.3390/app142311154 - 29 Nov 2024
Viewed by 719
Abstract
In orthodontics, the manual tracing of cephalometric radiographs is a common practice, where the Sella Turcica (ST) serves as a reference point. The radiologist often manually traces the outline of the sella using manual tools (e.g., calipers on radiographs). Perhaps the inherent complexity [...] Read more.
In orthodontics, the manual tracing of cephalometric radiographs is a common practice, where the Sella Turcica (ST) serves as a reference point. The radiologist often manually traces the outline of the sella using manual tools (e.g., calipers on radiographs). Perhaps the inherent complexity and variability in the shapes of sella and the lack of advanced assessment tools make the classification of sella challenging, as it requires extensive training, skills, time, and manpower to detect subtle changes that often may not be apparent. Moreover, existing semi-supervised learning (SSL) methods face key limitations such as shift invariance, inadequate feature representation, overfitting on small datasets, and a lack of generalization to unseen variations in ST morphology. Medical imaging data are often unlabeled, limiting the training of automated classification systems for ST morphology. To address these limitations, a novel semi-supervised deep subspace embedding (SSLDSE) framework is proposed. This approach integrates real-time stochastic augmentation to significantly expand the training dataset and introduce natural variability in the ST morphology, overcoming the constraints of small and non-representative datasets. Non-linear features are extracted and mapped to a non-linear subspace using Kullback–Leibler divergence, which ensures that the model remains consistent despite image transformations, thus resolving issues related to shift invariance. Additionally, fine-tuning the Inception-ResNet-v2 network on these enriched features reduces retraining costs when new unlabeled data becomes available. t-distributed stochastic neighbor embedding (t-SNE) is employed for effective feature representation through manifold learning, capturing complex patterns that previous methods might miss. Finally, a zero-shot classifier is utilized to accurately categorize the ST, addressing the challenge of classifying new or unseen variations. Further, the proposed SSLDSE framework is evaluated through comparative analysis with the existing methods (Active SSL, GAN SSL, Contrastive SSL, Modified Inception-ResNet-v2) for ST classification using various evaluation metrics. The SSLDSE and the existing methods are trained on our dataset (sourced from PGI Chandigarh, India), and a blind test is conducted on the benchmark dataset (IEEE ISBI 2015). The proposed method improves classification accuracy by 15% compared to state-of-the-art models and reduces retraining costs. Full article
(This article belongs to the Special Issue Application of Artificial Intelligence in Biomedical Informatics)
►▼ Show Figures

Figure 1

Figure 1
<p>Sample images of pre-defined Sella Turcica (ST) shapes: (<b>A</b>) Oval ST, (<b>B</b>) Circular ST, (<b>C</b>) Flat ST, and (<b>D</b>) Bridging ST. This study classified Circular ST as non-bridging, and Bridging ST was used for binary classification.</p> Full article ">Figure 2
<p>The schematic representation of a Hybrid Database (<span class="html-italic">L</span>) and Hybrid Case Base (<math display="inline"><semantics> <mrow> <mi>S</mi> <mi>L</mi> </mrow> </semantics></math>) from labeled (<math display="inline"><semantics> <mrow> <mi>S</mi> <msub> <mi>L</mi> <mi>i</mi> </msub> </mrow> </semantics></math>) and unlabeled (<math display="inline"><semantics> <mrow> <mi>S</mi> <msub> <mi>L</mi> <mover accent="true"> <mi>i</mi> <mo>^</mo> </mover> </msub> </mrow> </semantics></math>) case data. Feature extraction using KL divergence, mean (<math display="inline"><semantics> <mi>μ</mi> </semantics></math>), and standard deviation (<math display="inline"><semantics> <mi>σ</mi> </semantics></math>) is applied to both databases. Labeled data form a featured database with labels (<math display="inline"><semantics> <msub> <mi>L</mi> <mi>i</mi> </msub> </semantics></math>), while unlabeled data create a featured database without labels (<math display="inline"><semantics> <msub> <mi>L</mi> <mover accent="true"> <mi>i</mi> <mo>^</mo> </mover> </msub> </semantics></math>). A dynamic responsive data and label mechanism integrates both, resulting in (1) a Hybrid Database (<span class="html-italic">L</span>) and (2) a Hybrid Case Base (<math display="inline"><semantics> <mrow> <mi>S</mi> <mi>L</mi> </mrow> </semantics></math>) for further analysis.</p> Full article ">Figure 3
<p>Process flow diagram of the proposed SSLDSE framework.</p> Full article ">Figure 4
<p>The figure illustrates a comprehensive framework of the proposed SSLDSE that integrates labeled (<math display="inline"><semantics> <mrow> <mi>S</mi> <msub> <mi>L</mi> <mi>i</mi> </msub> </mrow> </semantics></math>) and unlabeled case databases (<math display="inline"><semantics> <mrow> <mi>S</mi> <msub> <mi>L</mi> <mover accent="true"> <mi>i</mi> <mo>^</mo> </mover> </msub> </mrow> </semantics></math>). Features are extracted using Kullback–Leibler divergence, mean (<math display="inline"><semantics> <mi>μ</mi> </semantics></math>), and standard deviation (<math display="inline"><semantics> <mi>σ</mi> </semantics></math>), forming a Hybrid Database. The data undergo stochastic augmentation and are processed through an Inception-ResNet-V2 model. A deep subspace descriptor with t-SNE refines the feature representations, and the outputs are classified by a zero-shot classifier (<math display="inline"><semantics> <mrow> <mi>Z</mi> <mi>s</mi> <mi>C</mi> </mrow> </semantics></math>) with KL divergence loss, enabling the model to handle unseen or unlabeled ST structures.</p> Full article ">Figure 5
<p>The illustrated SSLDSE architectural framework processes labeled (L) and semi-labeled (<math display="inline"><semantics> <mrow> <mi>S</mi> <msub> <mi>L</mi> <mi>i</mi> </msub> </mrow> </semantics></math>) data using Inception-ResNet-V2 as the CNN backbone to extract features (P, Q) and estimate pairwise probability densities (<math display="inline"><semantics> <msub> <mi mathvariant="normal">P</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> </semantics></math>, <math display="inline"><semantics> <msub> <mi mathvariant="normal">Q</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> </semantics></math>). KL divergence (<math display="inline"><semantics> <mrow> <msub> <mi>D</mi> <mi>KL</mi> </msub> <mrow> <mo>(</mo> <mi>P</mi> <mo>‖</mo> <mi>Q</mi> <mo>)</mo> </mrow> </mrow> </semantics></math>) minimizes divergence through the optimization of (Y). Manifold learning maps feature matrices (X) to t-SNE representations (Y) while preserving structural relationships. The SSL framework employs deep embedding and clustering (mean: <math display="inline"><semantics> <msub> <mi>μ</mi> <mi>j</mi> </msub> </semantics></math>, covariance: <math display="inline"><semantics> <msub> <mi>Σ</mi> <mi>j</mi> </msub> </semantics></math>) for feature representation. The zero-shot classifier constructs semantic vectors and applies KL divergence loss for output prediction (<math display="inline"><semantics> <msub> <mi>O</mi> <mi>i</mi> </msub> </semantics></math>).</p> Full article ">Figure 6
<p>Confusion matrix and ROC curve showcasing the validation results of the binary classifier, highlighting the proposed model’s classification performance through true positive/negative rates and the AUC-ROC score.</p> Full article ">Figure 7
<p>t-SNE plots visualizing the quantitative assessment of the proposed SSLDSE method. The plots illustrate the effective separation between bridging and non-bridging labels from our proprietary and IEEE ISBI 2015 datasets, demonstrating a clear class distinction.</p> Full article ">Figure 8
<p>Boxplots illustrate a detailed comparison of the classification error rates for the proposed SSLDSE method, showing the distribution of error rates across (<b>a</b>) the proprietary dataset and (<b>b</b>) the IEEE ISBI 2015 dataset, highlighting the variability and consistency in classification accuracy.</p> Full article ">Figure 9
<p>Boxplots comparing classification error rates across utilized SSL approaches and the proposed SSLDSE method, illustrating performance differences and the effectiveness of SSLDSE in reducing classification errors.</p> Full article ">Figure 10
<p>Visual interpretation of errors in ST-binary classification predictions, illustrating misclassified instances and highlighting the areas where the model’s predictions diverge from the true labels.</p> Full article ">
13 pages, 2433 KiB  
Article
Multi-Model Gait-Based KAM Prediction System Using LSTM-RNN and Wearable Devices
by Doyun Jung, Cheolwon Lee and Heung Seok Jeon
Appl. Sci. 2024, 14(22), 10721; https://doi.org/10.3390/app142210721 - 19 Nov 2024
Viewed by 752
Abstract
The purpose of this study is to develop an optimized system for predicting Knee Adduction Moment (KAM) using wearable Inertial Measurement Unit (IMU) sensors and Long Short-Term Memory (LSTM) RNN. Traditional KAM measurement methods are limited by the need for complex laboratory equipment [...] Read more.
The purpose of this study is to develop an optimized system for predicting Knee Adduction Moment (KAM) using wearable Inertial Measurement Unit (IMU) sensors and Long Short-Term Memory (LSTM) RNN. Traditional KAM measurement methods are limited by the need for complex laboratory equipment and significant time and cost investments. This study proposes two systems for predicting Knee Adduction Moment based on wearable IMU sensor data and gait patterns: the Multi-model Gait-based KAM Prediction System and the Single-model KAM Prediction System. The Multi-model system pre-classifies different gait patterns and uses specific prediction models tailored for each pattern, while the Single-model system handles all gait patterns with one unified model. Both systems were evaluated using IMU sensor data and GRF data collected from participants in a controlled laboratory environment. The overall performance of the Multi-model Gait-based KAM Prediction System showed an approximately 20% improvement over the Single-model KAM Prediction System. Specifically, the RMSE for the Multi-model system was 6.84 N·m, which is lower than the 8.82 N·m of the Single-model system, indicating a better predictive accuracy. The Multi-model system also achieved a MAPE of 8.47%, compared with 12.95% for the Single-model system, further demonstrating its superior performance. Full article
(This article belongs to the Special Issue Application of Artificial Intelligence in Biomedical Informatics)
►▼ Show Figures

Figure 1

Figure 1
<p>Overview of Multi-model Gait-based KAM Prediction System.</p> Full article ">Figure 2
<p>Laboratory environment with motion camera (<b>A</b>) and force plate (<b>B</b>) installed.</p> Full article ">Figure 3
<p>Detailed sensor attachment locations for the participants.</p> Full article ">Figure 4
<p>Collecting datasets with IMU and GRF sensors.</p> Full article ">Figure 5
<p>Organization of models in Multi-model Gait-based KAM Prediction System.</p> Full article ">Figure 6
<p>Confusion matrix of Multi-model Gait-based KAM Prediction System (Knee Thrust 97.15%, Normal 94.03%, Toe in 92.59%, Toe out 89.77%, Trunk lean 82.23%).</p> Full article ">Figure 7
<p>Comparison of the actual vs. predicted KAM values for the Multi-model and Single-model Gait-based KAM Prediction Systems. The close alignment of actual and predicted values in the Multi-model system demonstrates a higher prediction accuracy, especially across varying gait patterns.</p> Full article ">Figure 8
<p>Predictive accuracy visualization for the Multi-model and Single-model KAM Prediction Systems. Data points in the Multi-model system cluster more closely along the <math display="inline"><semantics> <mrow> <mi>y</mi> <mo>=</mo> <mi>x</mi> </mrow> </semantics></math> line, highlighting its enhanced accuracy and robustness compared with the Single-model system, particularly in managing diverse gait patterns.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-4
Back to TopTop