Journal of Medical and Biological Engineering最新文献

Purpose

The presence of motion artifacts (MA) in cardiac signals negatively impacts the reliability of higher-level information such as the Heart Rate (HR), and therefore the correct diagnosis of pathologies. This paper proposes an MA detection method, based on One-Dimensional Convolutional Neural Networks (1D CNN), to label noisy zones of signals as unreliable, and subsequently avoid them for metric calculations.

Methods

To validate the concept, we first design a CNN to detect MAs in electrocardiogram (ECG) recordings from MIT–BIH Arrhythmia and Noise Stress Test Databases. This network extracts features from 1 s data segments, and then classifies them as clean or noisy. Also, we then train a tuned version of the model with semi-synthetic ballistocardiogram (BCG) signals.

Results

The classification in ECG achieves an accuracy of 95.9% and the BCG classification obtains an accuracy of 91.1%. Both classifiers are incorporated into beat detection systems, which produce an increase in the sensitivity of the detection algorithms from 75 to 98.5% in the ECG case, and from 72.1 to 94.5% in the case of BCG, for signals contaminated at 0 dB of SNR.

Conclusion

We propose that this method will improve accuracy of any processing algorithm on BCG signals by identifying useful segments where a high accuracy can be achieved.

Purpose

In this paper, we propose an open-source deep learning-based computer-aided diagnosis system for breast ultrasound images based on the Breast Imaging Reporting and Data System (BI-RADS).

Methods

Our dataset with 8,026 region-of-interest images preprocessed with ten times data augmentation. We compared the classification performance of VGG-16, ResNet-50, and DenseNet-121 and two ensemble methods integrated the single models.

Results

The ensemble model achieved the best performance, with 81.8% accuracy. Our results show that our model is performant enough to classify Category 2 and Category 4/5 lesions, and data augmentation can improve the classification performance of Category 3.

Conclusion

Our main contribution is to classify breast ultrasound lesions into BI-RADS assessment classes that place more emphasis on adhering to the BI-RADS medical suggestions including recommending routine follow-up tracing (Category 2), short-term follow-up tracing (Category 3) and biopsies (Category 4/5).

Purpose

Segmentation of nuclei and cytoplasm in cellular images is essential for estimating the prognosis of lung cancer disease. The detection of these organelles in the unstained brightfield microscopic images is challenging due to poor contrast and lack of separation of structures with irregular morphology. This work aims to carry out semantic segmentation of nuclei and cytoplasm in lung cancer brightfield images using the Swin-Unet Transformer.

Methods

For this study, publicly available brightfield images of lung cancer cells are pre-processed and fed to the Swin-Unet for semantic segmentation. Model specific hyperparameters are identified after detailed analysis and the segmentation performance is validated using standard evaluation metrics.

Results

The hyperparameter analysis provides the selection of optimum parameters as focal loss, learning rate of 0.0001, Adam optimizer, and Swin Transformer patch size of 4. The obtained results show that with these parameters, the Swin-Unet Transformer accurately segmented the nuclei and cytoplasm in the brightfield images with pixel-F1 scores of 90.71% and 79.29% respectively.

Conclusion

It is observed that the model could identify nuclei and cytoplasm with varied morphologies. The detection of cytoplasm with weak and subtle edge details indicates the effectiveness of shifted window based self attention mechanism of Swin-Unet in capturing the global and long distance pixel interactions in the brightfield images. Thus, the adopted methodology in this study can be employed for the precise segmentation of nuclei and cytoplasm for assessing the malignancy of lung cancer disease.

Purpose

This paper aims to analyze the influence of mechanical force on bone regeneration from macro and micro perspectives, to investigate the mechanical response of bone tissues at various scales after operation and provide a theoretical basis for further research and clinical practice.

Methods

An effective postoperative lumbar model was constructed, and the bone regeneration area was established at the osteotomy. The area was divided into five stages, from 10 MPa to 100 MPa. Then, the osteon and bone lacuna-osteocyte models were constructed, and their biomechanical characteristics under different working conditions were studied.

Results

From the first stage to the fifth stage, the macroscopic bone tissue larger than 3000 µε decreased by about 40%, the maximum stress ratio n approximates k (E_O/E_T) of macro- and micro-bone tissues, and the area of osteocytes less than 3000 µε increased by about 45%. In the second stage, 41.7% of the bone cells have a strain of 1000 µε ∼ 3000 µε, and this percentage increases to 66.7%∼72.2% after the fourth stage.

Conclusion

The macro-meso stress ratio is related to the tissue strength around the osteon. In the first stage, the patient should lie flat and rest, instead of standing upright. At the beginning of the fourth stage, the rate of bone regeneration is much faster than the rate of lesions, making it suitable for upright recovery, and the recovery speed increases.

Purpose

Mammography is the modality of choice for the early detection of breast cancer. Deep learning, using convolutional neural networks (CNNs) specifically, have achieved extraordinary results in the classification of diseases, including breast cancer, on imaging. The images used to train a CNN varies based on several factors, such as imaging technique, imaging equipment, and study population; these factors significantly affect the accuracy of the CNN models. The aim of this study was to develop a novel CNN for the classification of mammograms as benign or malignant and to compare its utility to that of popular pre-trained CNNs in the literature using transfer learning. All CNNs were trained to detect breast cancer on mammograms using mammograms from a created database of Mexican women (MAMMOMX-PABIOM) and from a public database of UK women (MIAS).

Methods

A database (MAMMOMX-PABIOM) was built comprising 1,070 mammography images of 235 Mexican patients from 4 hospitals in Mexico. The study also used mammographic images from the Mammographic Image Analysis Society (MIAS) public database, which comprises mammography images from the UK National Breast Screening Programme. A novel CNN was developed and trained based on different configurations of training data; the accuracy of the models resulting from the novel CNN were compared with models resulting from more advanced pre-trained CNNs (DenseNet121, MobileNetV2, ResNet 50, VGG16) which were built using transfer learning.

Results

Of the models resulting from pre-trained CNNs using transfer learning, the model based on MobileNetV2 and training data from the MAMMOMX-PABIOM database achieved the highest validation accuracy of 70.10%. In comparison, the novel CNN, when trained with the data configuration A6, which comprises data from both the MAMMOMX-PABIOM database and the MIAS database, produced a much higher accuracy of 99.14%.

Conclusion

Although transfer learning is a widely used technique when training, data is scarce. The novel CNN produced much higher accuracy values across all configurations of training data compared to the accuracy values of pre-trained CNNs using transfer learning. In addition, this study addresses the gap in that neither a national database of mammograms of Mexican women exists, nor a deep learning tool for the classification of mammograms as benign or malignant that is focused on this population.

Purpose

Pancreatic ductal adenocarcinoma (PDAC) is the most prevalent form of pancreatic cancer, accounting for about 85% of all occurrences. It is highly challenging to treat PDAC because of its extreme aggressiveness and lack of therapeutic options. Identifying new gene markers can help in the design of novel targeted therapeutics.

Methods

In this study, we identified three different gene prognostic markers in PDAC using a machine learning approach. Initially, the differential expression genes (DEGs) profile of accession number GSE183795 was downloaded from the gene expression omnibus database of the National Center for Biotechnology Information (NCBI), which consists of the expression profile of the 244 patients with PDAC (139 pancreatic tumors, 102 adjacent non-tumors and 3 normal). Then, the expression dataset was preprocessed using different packages of R programming, such as GEOquery, Affy, and Limma. Further, DEGs were identified by the machine learning algorithms, including random forest (RF) and extreme gradient boost (XGboost). Finally, survival analysis was performed to identify DEGs using GEPIA software (TCGA database).

Results

Our results revealed that 6 out of 25 DEGs (ERCC3, ACY3, ATP2A3, MW-TW1879, MW-TW3829, and ZBTB7A) identified by RF and XGBoost algorithm were the same, indicating their feature importance. Moreover, three genes, including ATP2A3 (p = 0.029), NRL (p = 0.012), and FBXO45 (p = 0.013), were statistically significant when tested for survival analysis and may be utilized as the prognostic marker genes for PDAC.

Conclusion

These findings provide valuable insights into the molecular characteristics of PDAC and can potentially guide future research on cancer theranostics interventions for this devastating disease.

Purpose

A brain tumor is defined as any group of atypical cells occupying space in the brain. There are more than 120 types of them. MRI scans are used for brain tumor diagnosis since they are more detailed and three-dimensional. Accurate localization and segmentation of the tumor portion increase the patients' survival rates. To this end, we presented a systematic review of the latest development of brain tumor segmentation from MRI.

Methods

To find related articles, we searched the keywords like "brain tumors" and "segmentation by MRI”. The searches were performed on Elsevier, Springer, Wiley, and the leading conferences in the field of medical image processing. A total of 79 publications dedicated to tumor segmentation from years 2019 to 2023 were selected and categorized into four categories: non-Artificial Intelligence, machine learning, deep learning, and hybrid deep learning methods.

Results

We reviewed the trending techniques of tumor segmentation and provided a unified and integrated overview of the current state-of-the-art. The article dealt with providing the capabilities and shortcomings associated with each approach and the restrictions on using automated medical image segmentation techniques in clinical practice were determined.

Conclusion

In this study, the advancement of brain tumor segmentation by MRI is discussed, focusing more on recent articles. It identified the restrictions of the presented techniques regarding the four mentioned categories, which prevent them from being used in clinical practice. The literature will guide the researchers to become familiar with both the leading techniques and the potential problems that need to be addressed.

Purpose

This research focuses on developing new ways to monitor coronary artery disease (CAD), the leading type of cardiovascular disease, which requires more straightforward, safer, and more continuous tracking methods along with angiography, the gold standard method. This need arises due to the high risk, cost, and the large number of people living with undiagnosed CAD. The study explores the use of the intrinsic frequency (IF) method, a promising but underutilized technique in the realm of CAD monitoring, to investigate its effectiveness in identifying CAD through the analysis of radial pressure wave patterns.

Method

The radial pressure waves, alongside major CAD risk factors (hypertension, diabetes, hyperlipidemia, smoking, family history, age, and sex) were analyzed in 100 patients undergoing angiography. The IF method was utilized to evaluate the dynamics of heart and arterial system function, focusing on specific IF indices that reflect vasculature health level.

Result

The results, validated through T-tests, reveal notable alterations in specific IF indices among CAD patients: ({{varvec{omega}}}_{2}) shows a significant increase with a mean of 82.5 bpm in CAD versus 41.56 bpm in non-CAD cases. Similarly, ({varvec{Delta}}{varvec{omega}}) displays a significant decrease with a mean of 15.73 bpm in CAD compared to 49.02 bpm in non-CAD individuals. Conversely, ({{varvec{omega}}}_{1}) demonstrates minimal variance between CAD and non-CAD groups.

Conclusion

This study underscores the potential of IF indices, particularly ({{varvec{omega}}}_{2}) and ({varvec{Delta}}{varvec{omega}}), as markers for severe CAD cases and strongly advocate for the integration of continuous monitoring strategies via modern technology in healthcare, such as smartwatches in CAD management.

Graphical Abstract

Purpose

This review offers insight into AI’s current and future contributions to medical image analysis. The article highlights the challenges associated with manual image interpretation and introduces AI methodologies, including machine learning and deep learning. It explores AI’s applications in image segmentation, classification, registration, and reconstruction across various modalities like X-ray, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound.

Background

Medical image analysis is vital in modern healthcare, facilitating disease diagnosis, treatment, and monitoring. Integrating artificial intelligence (AI) techniques, particularly deep learning, has revolutionized this field.

Methods

Recent advancements are discussed, such as generative adversarial networks (GANs), transfer learning, and federated learning. The review assesses the advantages and limitations of AI in medical image analysis, underscoring the importance of interpretability, robustness, and generalizability in clinical practice. Ethical considerations related to data privacy, bias, and regulatory aspects are also examined.

Results

The article concludes by exploring future directions, including personalized medicine, multi-modal fusion, real-time analysis, and seamless integration with electronic health records (EHRs).

Conclusion

This comprehensive review delineates artificial intelligence’s current and prospective role in medical image analysis. With implications for researchers, clinicians, and policymakers, it underscores AI’s transformative potential in enhancing patient care.

Purpose

Immediate hypersensitivity reactions, commonly triggered by allergens, play a crucial role in clinical allergies. The skin prick test is the primary diagnostic tool for allergy, involving the application of an allergen drop on the forearm's volar surface. A sterile lancet is then used to cross the drop, observing the formation of a wheal if sensitized. In allergy practice, wheals are quantified using an arbitrary visual scale or methods such as the Dermographic Pen Method, involving a dermographic pen and graph paper, or a centimeter ruler. These methodologies are semi-quantitative, time-consuming, and operator-dependent. This study addresses the need for accurate and standardized quantification of SPT responses. We developed a Semi-Automated Method (SAM) for wheal detection to achieve this.

Methods

A cohort of 26 patients with respiratory allergies underwent SPTs with various allergens. Wheals were quantified using three methods: Arbitrary Visual Scale (AVSM), DPMM (Dermographic Pen Measurement Method), and the newly developed SAM. SAM utilized photographic detection and image analysis, and calculated major and minor diameters, mean diameter, wheal surface area, and skin index.

Results

Comparative analysis revealed SAM's superior performance in precision and efficiency compared to AVSM and DPMM. Mean surface measurements of histamine-generated wheals using SAM were significantly lower than those obtained with DPMM. Interestingly, SAM consistently demonstrated better performance across all tested allergens.

Conclusion

The introduction of SAM represents a significant advancement in allergy diagnostics. Its semi-automated approach enhances precision and facilitates long-term monitoring of SPT results. Through automation, SAM achieves accuracy in results and ease of use, notably improving allergy diagnostics.