Pub Date : 2025-01-21DOI: 10.1088/2057-1976/ada8b0
Oscar Jalnefjord, Nicolas Geades, Guillaume Gilbert, Isabella M Björkman-Burtscher, Maria Ljungberg
Dual-polarity readout is a simple and robust way to mitigate Nyquist ghosting in diffusion-weighted echo-planar imaging but imposes doubled scan time. We here propose how dual-polarity readout can be implemented with little or no increase in scan time by exploiting an observed b-value dependence and signal averaging. The b-value dependence was confirmed in healthy volunteers with distinct ghosting at low b-values but of negligible magnitude atb= 1000 s/mm2. The usefulness of the suggested strategy was exemplified with a scan using tensor-valued diffusion encoding for estimation of parameter maps of mean diffusivity, and anisotropic and isotropic mean kurtosis, showing that ghosting propagated into all three parameter maps unless dual-polarity readout was applied. Results thus imply that extending the use of dual-polarity readout to low non-zero b-values provides effective ghost elimination and can be used without increased scan time for any diffusion MRI scan containing signal averaging at low b-values.
{"title":"Nyquist ghost elimination for diffusion MRI by dual-polarity readout at low b-values.","authors":"Oscar Jalnefjord, Nicolas Geades, Guillaume Gilbert, Isabella M Björkman-Burtscher, Maria Ljungberg","doi":"10.1088/2057-1976/ada8b0","DOIUrl":"10.1088/2057-1976/ada8b0","url":null,"abstract":"<p><p>Dual-polarity readout is a simple and robust way to mitigate Nyquist ghosting in diffusion-weighted echo-planar imaging but imposes doubled scan time. We here propose how dual-polarity readout can be implemented with little or no increase in scan time by exploiting an observed b-value dependence and signal averaging. The b-value dependence was confirmed in healthy volunteers with distinct ghosting at low b-values but of negligible magnitude at<i>b</i>= 1000 s/mm<sup>2</sup>. The usefulness of the suggested strategy was exemplified with a scan using tensor-valued diffusion encoding for estimation of parameter maps of mean diffusivity, and anisotropic and isotropic mean kurtosis, showing that ghosting propagated into all three parameter maps unless dual-polarity readout was applied. Results thus imply that extending the use of dual-polarity readout to low non-zero b-values provides effective ghost elimination and can be used without increased scan time for any diffusion MRI scan containing signal averaging at low b-values.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142962166","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-21DOI: 10.1088/2057-1976/ada8ad
Francisco Fumero, Jose Sigut, José Estévez, Tinguaro Díaz-Alemán
This paper systematically evaluates saliency methods as explainability tools for convolutional neural networks trained to diagnose glaucoma using simplified eye fundus images that contain only disc and cup outlines. These simplified images, a methodological novelty, were used to relate features highlighted in the saliency maps to the geometrical clues that experts consider in glaucoma diagnosis. Despite their simplicity, these images retained sufficient information for accurate classification, with balanced accuracies ranging from 0.8331 to 0.8890, compared to 0.8090 to 0.9203 for networks trained on the original images. The study used a dataset of 606 images, along with RIM-ONE DL and REFUGE datasets, and explored nine saliency methods. A discretization algorithm was applied to reduce noise and compute normalized attribution values for standard eye fundus sectors. Consistent with other medical imaging studies, significant variability was found in the attribution maps, influenced by the method, model, or architecture, and often deviating from typical sectors experts examine. However, globally, the results were relatively stable, with a strong correlation of 0.9289 (p < 0.001) between relevant sectors in our dataset and RIM-ONE DL, and 0.7806 (p < 0.001) for REFUGE. The findings suggest caution when using saliency methods in critical fields like medicine. These methods may be more suitable for broad image relevance interpretation rather than assessing individual cases, where results are highly sensitive to methodological choices. Moreover, the regions identified by the networks do not consistently align with established medical criteria for disease severity.
{"title":"Systematic application of saliency maps to explain the decisions of convolutional neural networks for glaucoma diagnosis based on disc and cup geometry.","authors":"Francisco Fumero, Jose Sigut, José Estévez, Tinguaro Díaz-Alemán","doi":"10.1088/2057-1976/ada8ad","DOIUrl":"10.1088/2057-1976/ada8ad","url":null,"abstract":"<p><p>This paper systematically evaluates saliency methods as explainability tools for convolutional neural networks trained to diagnose glaucoma using simplified eye fundus images that contain only disc and cup outlines. These simplified images, a methodological novelty, were used to relate features highlighted in the saliency maps to the geometrical clues that experts consider in glaucoma diagnosis. Despite their simplicity, these images retained sufficient information for accurate classification, with balanced accuracies ranging from 0.8331 to 0.8890, compared to 0.8090 to 0.9203 for networks trained on the original images. The study used a dataset of 606 images, along with RIM-ONE DL and REFUGE datasets, and explored nine saliency methods. A discretization algorithm was applied to reduce noise and compute normalized attribution values for standard eye fundus sectors. Consistent with other medical imaging studies, significant variability was found in the attribution maps, influenced by the method, model, or architecture, and often deviating from typical sectors experts examine. However, globally, the results were relatively stable, with a strong correlation of 0.9289 (<i>p</i> < 0.001) between relevant sectors in our dataset and RIM-ONE DL, and 0.7806 (<i>p</i> < 0.001) for REFUGE. The findings suggest caution when using saliency methods in critical fields like medicine. These methods may be more suitable for broad image relevance interpretation rather than assessing individual cases, where results are highly sensitive to methodological choices. Moreover, the regions identified by the networks do not consistently align with established medical criteria for disease severity.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142962167","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-20DOI: 10.1088/2057-1976/adabea
Abir Chaari, Imen Fourati Kallel, Sonda Kammoun, Mondher Frikha
Deep learning has emerged as a powerful tool in medical imaging, particularly for corneal topographic map classification. However, the scarcity of labeled data poses a significant challenge to achieving robust performance. This study investigates the impact of various data augmentation strategies on enhancing the performance of a customized convolutional neural network model for corneal topographic map classification. We propose a hybrid data augmentation approach that combines traditional transformations, generative adversarial networks, and specific generative models. Experimental results demonstrate that the hybrid data augmentation method, achieves the highest accuracy of 99.54%, significantly outperforming individual data augmentation techniques. This hybrid approach not only improves model accuracy but also mitigates overfitting issues, making it a promising solution for medical image classification tasks with limited data availability.
{"title":"Hybrid Data Augmentation Strategies for Robust Deep Learning Classification of Corneal Topographic MapTopographic Map.","authors":"Abir Chaari, Imen Fourati Kallel, Sonda Kammoun, Mondher Frikha","doi":"10.1088/2057-1976/adabea","DOIUrl":"https://doi.org/10.1088/2057-1976/adabea","url":null,"abstract":"<p><p>Deep learning has emerged as a powerful tool in medical imaging, particularly for corneal topographic map classification. However, the scarcity of labeled data poses a significant challenge to achieving robust performance. This study investigates the impact of various data augmentation strategies on enhancing the performance of a customized convolutional neural network model for corneal topographic map classification. We propose a hybrid data augmentation approach that combines traditional transformations, generative adversarial networks, and specific generative models. Experimental results demonstrate that the hybrid data augmentation method, achieves the highest accuracy of 99.54%, significantly outperforming individual data augmentation techniques. This hybrid approach not only improves model accuracy but also mitigates overfitting issues, making it a promising solution for medical image classification tasks with limited data availability.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142999520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-20DOI: 10.1088/2057-1976/adabeb
Chase Haddix, Madison Bates, Sarah Garcia Pava, Elizabeth Salmon Powell, Lumy Sawaki, Sridhar Sunderam
Brain-computer interfaces (BCIs) offer disabled individuals the means to interact with devices by decoding the electroencephalogram (EEG). However, decoding intent in fine motor tasks can be challenging, especially in stroke survivors with cortical lesions. Here, we attempt to decode graded finger extension from the EEG in stroke patients with left-hand paresis and healthy controls. Participants extended their fingers to one of four levels: low, medium, high, or "no-go" (none), while hand, muscle (electromyography: EMG), and brain (EEG) activity were monitored. Event-related desynchronization (ERD) was measured as the change in 8-30 Hz EEG power during movement. Classifiers were trained on the ERD, EMG power, or both (EEG+EMG) to decode finger extension, and accuracy assessed via four-fold cross-validation for each hand of each participant. Mean accuracy exceeded chance (25%) for controls (n=11) at 62% for EMG, 60% for EEG, and 71% for EEG+EMG on the left hand; and 67%, 60%, and 74%, respectively, on the right hand. Accuracies were similar on the unimpaired right hand for the stroke group (n=3): 61%, 68%, and 78%, respectively. But on the paretic left hand, EMG only discriminated no-go from movement above chance (41%); in contrast, EEG gave 65% accuracy (68% for EEG+EMG), comparable to the non-paretic hand. The median ERD was significant (p < 0.01) over the cortical hand area in both groups and increased with each level of finger extension. But while the ERD favored the hemisphere contralateral to the active hand as expected, it was ipsilateral for the left hand of stroke due to the lesion in the right hemisphere, which may explain its discriminative ability. Hence, the ERD captures effort in finger extension regardless of success or failure at the task; and harnessing residual EMG improves the correlation. This marker could be leveraged in rehabilitative protocols that focus on fine motor control.
{"title":"Electroencephalogram Features Reflect Effort Corresponding to Graded Finger Extension: Implications for Hemiparetic Stroke.","authors":"Chase Haddix, Madison Bates, Sarah Garcia Pava, Elizabeth Salmon Powell, Lumy Sawaki, Sridhar Sunderam","doi":"10.1088/2057-1976/adabeb","DOIUrl":"https://doi.org/10.1088/2057-1976/adabeb","url":null,"abstract":"<p><p>Brain-computer interfaces (BCIs) offer disabled individuals the means to interact with devices by decoding the electroencephalogram (EEG). However, decoding intent in fine motor tasks can be challenging, especially in stroke survivors with cortical lesions. Here, we attempt to decode graded finger extension from the EEG in stroke patients with left-hand paresis and healthy controls. Participants extended their fingers to one of four levels: low, medium, high, or \"no-go\" (none), while hand, muscle (electromyography: EMG), and brain (EEG) activity were monitored. Event-related desynchronization (ERD) was measured as the change in 8-30 Hz EEG power during movement. Classifiers were trained on the ERD, EMG power, or both (EEG+EMG) to decode finger extension, and accuracy assessed via four-fold cross-validation for each hand of each participant. Mean accuracy exceeded chance (25%) for controls (n=11) at 62% for EMG, 60% for EEG, and 71% for EEG+EMG on the left hand; and 67%, 60%, and 74%, respectively, on the right hand. Accuracies were similar on the unimpaired right hand for the stroke group (n=3): 61%, 68%, and 78%, respectively. But on the paretic left hand, EMG only discriminated no-go from movement above chance (41%); in contrast, EEG gave 65% accuracy (68% for EEG+EMG), comparable to the non-paretic hand. The median ERD was significant (p < 0.01) over the cortical hand area in both groups and increased with each level of finger extension. But while the ERD favored the hemisphere contralateral to the active hand as expected, it was ipsilateral for the left hand of stroke due to the lesion in the right hemisphere, which may explain its discriminative ability. Hence, the ERD captures effort in finger extension regardless of success or failure at the task; and harnessing residual EMG improves the correlation. This marker could be leveraged in rehabilitative protocols that focus on fine motor control.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142999518","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Albumin and γ-globulin concentrations in subcutaneous adipose tissues (SAT) have been quantified by multivariate regression based on admittance relaxation time distribution (mraRTD) under the fluctuated background of sodium electrolyte concentration. The mraRTD formulates P = Ac + Ξ (P: peak matrix of distribution function magnitude ɣP and frequency τP, c: concentration matrix of albumin cAlb, γ-globulin Gloc, and sodium electrolyte Nac, A: coefficient matrix of a multivariate regression model, and Ξ: error matrix). The mraRTD is implemented by two processes which are: 1) the training process of A through the maximum likelihood estimation of P and 2) the quantification process of cAlb, Gloc, and Nac through the model prediction. In the training process, a positive correlation is present between cAlb, Gloc, and Nac to ɣP1 at τP1= 0.1 as well as ɣP2 at τP2= 1.40 µs as under a fixed concentration of proteins solution into a porcine SAT (cAlb = 0.800-2.400 g/dL, Gloc = 0.400-1.200 g/dL and Nac = 0.700-0.750 g/dL). The mraRTD method quantifies cAlb, Gloc, and Nac in SAT with an absolute error of 33.79%, 44.60%, and 2.18%, respectively.
{"title":"Quantification of Albumin and ɣ-Globulin Concentrations by Multivariate Regression Based on Admittance Relaxation Time Distribution (mrARTD).","authors":"Arbariyanto Mahmud Wicaksono, Daisuke Kawashima, Ryoma Ogawa, Shinsuke Akita, Masahiro Takei","doi":"10.1088/2057-1976/adabec","DOIUrl":"https://doi.org/10.1088/2057-1976/adabec","url":null,"abstract":"<p><p>Albumin and γ-globulin concentrations in subcutaneous adipose tissues (SAT) have been quantified by multivariate regression based on admittance relaxation time distribution (mraRTD) under the fluctuated background of sodium electrolyte concentration. The mraRTD formulates P = Ac + Ξ (P: peak matrix of distribution function magnitude ɣP and frequency τP, c: concentration matrix of albumin cAlb, γ-globulin Gloc, and sodium electrolyte Nac, A: coefficient matrix of a multivariate regression model, and Ξ: error matrix). The mraRTD is implemented by two processes which are: 1) the training process of A through the maximum likelihood estimation of P and 2) the quantification process of cAlb, Gloc, and Nac through the model prediction. In the training process, a positive correlation is present between cAlb, Gloc, and Nac to ɣP1 at τP1= 0.1 as well as ɣP2 at τP2= 1.40 µs as under a fixed concentration of proteins solution into a porcine SAT (cAlb = 0.800-2.400 g/dL, Gloc = 0.400-1.200 g/dL and Nac = 0.700-0.750 g/dL). The mraRTD method quantifies cAlb, Gloc, and Nac in SAT with an absolute error of 33.79%, 44.60%, and 2.18%, respectively.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142999522","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-17DOI: 10.1088/2057-1976/ada6ba
S A Yoganathan, Tarraf Torfeh, Satheesh Paloor, Rabih Hammoud, Noora Al-Hammadi, Rui Zhang
Backgroundand Purpose:This study aimed to develop and evaluate an efficient method to automatically segment T1- and T2-weighted brain magnetic resonance imaging (MRI) images. We specifically compared the segmentation performance of individual convolutional neural network (CNN) models against an ensemble approach to advance the accuracy of MRI-guided radiotherapy (RT) planning.Materials and Methods. The evaluation was conducted on a private clinical dataset and a publicly available dataset (HaN-Seg). Anonymized MRI data from 55 brain cancer patients, including T1-weighted, T1-weighted with contrast, and T2-weighted images, were used in the clinical dataset. We employed an EDL strategy that integrated five independently trained 2D neural networks, each tailored for precise segmentation of tumors and organs at risk (OARs) in the MRI scans. Class probabilities were obtained by averaging the final layer activations (Softmax outputs) from the five networks using a weighted-average method, which were then converted into discrete labels. Segmentation performance was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance at 95% (HD95). The EDL model was also tested on the HaN-Seg public dataset for comparison.Results. The EDL model demonstrated superior segmentation performance on both the clinical and public datasets. For the clinical dataset, the ensemble approach achieved an average DSC of 0.7 ± 0.2 and HD95 of 4.5 ± 2.5 mm across all segmentations, significantly outperforming individual networks which yielded DSC values ≤0.6 and HD95 values ≥14 mm. Similar improvements were observed in the HaN-Seg public dataset.Conclusions. Our study shows that the EDL model consistently outperforms individual CNN networks in both clinical and public datasets, demonstrating the potential of ensemble learning to enhance segmentation accuracy. These findings underscore the value of the EDL approach for clinical applications, particularly in MRI-guided RT planning.
{"title":"Automatic segmentation of MRI images for brain radiotherapy planning using deep ensemble learning.","authors":"S A Yoganathan, Tarraf Torfeh, Satheesh Paloor, Rabih Hammoud, Noora Al-Hammadi, Rui Zhang","doi":"10.1088/2057-1976/ada6ba","DOIUrl":"https://doi.org/10.1088/2057-1976/ada6ba","url":null,"abstract":"<p><p><i>Background</i><i>and Purpose</i><b>:</b>This study aimed to develop and evaluate an efficient method to automatically segment T1- and T2-weighted brain magnetic resonance imaging (MRI) images. We specifically compared the segmentation performance of individual convolutional neural network (CNN) models against an ensemble approach to advance the accuracy of MRI-guided radiotherapy (RT) planning.<i>Materials and Methods</i>. The evaluation was conducted on a private clinical dataset and a publicly available dataset (HaN-Seg). Anonymized MRI data from 55 brain cancer patients, including T1-weighted, T1-weighted with contrast, and T2-weighted images, were used in the clinical dataset. We employed an EDL strategy that integrated five independently trained 2D neural networks, each tailored for precise segmentation of tumors and organs at risk (OARs) in the MRI scans. Class probabilities were obtained by averaging the final layer activations (Softmax outputs) from the five networks using a weighted-average method, which were then converted into discrete labels. Segmentation performance was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance at 95% (HD95). The EDL model was also tested on the HaN-Seg public dataset for comparison.<i>Results</i>. The EDL model demonstrated superior segmentation performance on both the clinical and public datasets. For the clinical dataset, the ensemble approach achieved an average DSC of 0.7 ± 0.2 and HD95 of 4.5 ± 2.5 mm across all segmentations, significantly outperforming individual networks which yielded DSC values ≤0.6 and HD95 values ≥14 mm. Similar improvements were observed in the HaN-Seg public dataset.<i>Conclusions</i>. Our study shows that the EDL model consistently outperforms individual CNN networks in both clinical and public datasets, demonstrating the potential of ensemble learning to enhance segmentation accuracy. These findings underscore the value of the EDL approach for clinical applications, particularly in MRI-guided RT planning.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":"11 2","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142999526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-17DOI: 10.1088/2057-1976/ada6bc
Loganathan R, Latha S
Global blindness is substantially influenced by age-related macular degeneration (AMD). It significantly shortens people's lives and severely impairs their visual acuity. AMD is becoming more common, requiring improved diagnostic and prognostic methods. Treatment efficacy and patient survival rates stand to benefit greatly from these upgrades. To improve AMD diagnosis in preprocessed retinal images, this study uses Grey Level Co-occurrence Matrix (GLCM) features for texture analysis. The selected GLCM features include contrast and dissimilarity. Notably, grayscale pixel values were also integrated into the analysis. Key factors such as contrast, correlation, energy, and homogeneity were identified as the primary focuses of the study. Various supervised machine learning (ML) and CNN techniques were employed on Optical Coherence Tomography (OCT) image datasets. The impact of feature selection on model performance is evaluated by comparing all GLCM features, selected GLCM features, and grayscale pixel features. Models using GSF features showed low accuracy, with OCTID at 23% and Kermany at 54% for BC, and 23% and 53% for CNN. In contrast, GLCM features achieved 98% for OCTID and 73% for Kermany in RF, and 83% and 77% in CNN. SFGLCM features performed the best, achieving 98% for OCTID across both RF and CNN, and 77% for Kermany. Overall, SFGLCM and GLCM features outperformed GSF, improving accuracy, generalization, and reducing overfitting for AMD detection. The Python-based research demonstrates ML's potential in ophthalmology to enhance patient outcomes.
{"title":"Enhanced AMD detection in OCT images using GLCM texture features with Machine Learning and CNN methods.","authors":"Loganathan R, Latha S","doi":"10.1088/2057-1976/ada6bc","DOIUrl":"10.1088/2057-1976/ada6bc","url":null,"abstract":"<p><p>Global blindness is substantially influenced by age-related macular degeneration (AMD). It significantly shortens people's lives and severely impairs their visual acuity. AMD is becoming more common, requiring improved diagnostic and prognostic methods. Treatment efficacy and patient survival rates stand to benefit greatly from these upgrades. To improve AMD diagnosis in preprocessed retinal images, this study uses Grey Level Co-occurrence Matrix (GLCM) features for texture analysis. The selected GLCM features include contrast and dissimilarity. Notably, grayscale pixel values were also integrated into the analysis. Key factors such as contrast, correlation, energy, and homogeneity were identified as the primary focuses of the study. Various supervised machine learning (ML) and CNN techniques were employed on Optical Coherence Tomography (OCT) image datasets. The impact of feature selection on model performance is evaluated by comparing all GLCM features, selected GLCM features, and grayscale pixel features. Models using GSF features showed low accuracy, with OCTID at 23% and Kermany at 54% for BC, and 23% and 53% for CNN. In contrast, GLCM features achieved 98% for OCTID and 73% for Kermany in RF, and 83% and 77% in CNN. SFGLCM features performed the best, achieving 98% for OCTID across both RF and CNN, and 77% for Kermany. Overall, SFGLCM and GLCM features outperformed GSF, improving accuracy, generalization, and reducing overfitting for AMD detection. The Python-based research demonstrates ML's potential in ophthalmology to enhance patient outcomes.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142943746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-15DOI: 10.1088/2057-1976/ada6bd
Sho Maruyama, Fumiya Mizutani, Haruyuki Watanabe
Objectives:In digital image diagnosis using medical displays, it is crucial to rigorously manage display devices to ensure appropriate image quality and diagnostic safety. The aim of this study was to develop a model for the efficient quality control (QC) of medical displays, specifically addressing the measurement items of contrast response and maximum luminance as part of constancy testing, and to evaluate its performance. In addition, the study focused on whether these tasks could be addressed using a multitasking strategy.Methods:The model used in this study was constructed by fine-tuning a pretrained model and expanding it to a multioutput configuration that could perform both contrast response classification and maximum luminance regression. QC images displayed on a medical display were captured using a smartphone, and these images served as the input for the model. The performance was evaluated using the area under the receiver operating characteristic curve (AUC) for the classification task. For the regression task, correlation coefficients and Bland-Altman analysis were applied. We investigated the impact of different architectures and verified the performance of multi-task models against single-task models as a baseline.Results:Overall, the classification task achieved a high AUC of approximately 0.9. The correlation coefficients for the regression tasks ranged between 0.6 and 0.7 on average. Although the model tended to underestimate the maximum luminance values, the error margin was consistently within 5% for all conditions.Conclusion:These results demonstrate the feasibility of implementing an efficient QC system for medical displays and the usefulness of a multitask-based method. Thus, this study provides valuable insights into the potential to reduce the workload associated with medical-device management the development of QC systems for medical devices, highlighting the importance of future efforts to improve their accuracy and applicability.
{"title":"Novel approach for quality control testing of medical displays using deep learning technology.","authors":"Sho Maruyama, Fumiya Mizutani, Haruyuki Watanabe","doi":"10.1088/2057-1976/ada6bd","DOIUrl":"10.1088/2057-1976/ada6bd","url":null,"abstract":"<p><p><i>Objectives:</i>In digital image diagnosis using medical displays, it is crucial to rigorously manage display devices to ensure appropriate image quality and diagnostic safety. The aim of this study was to develop a model for the efficient quality control (QC) of medical displays, specifically addressing the measurement items of contrast response and maximum luminance as part of constancy testing, and to evaluate its performance. In addition, the study focused on whether these tasks could be addressed using a multitasking strategy.<i>Methods:</i>The model used in this study was constructed by fine-tuning a pretrained model and expanding it to a multioutput configuration that could perform both contrast response classification and maximum luminance regression. QC images displayed on a medical display were captured using a smartphone, and these images served as the input for the model. The performance was evaluated using the area under the receiver operating characteristic curve (AUC) for the classification task. For the regression task, correlation coefficients and Bland-Altman analysis were applied. We investigated the impact of different architectures and verified the performance of multi-task models against single-task models as a baseline.<i>Results:</i>Overall, the classification task achieved a high AUC of approximately 0.9. The correlation coefficients for the regression tasks ranged between 0.6 and 0.7 on average. Although the model tended to underestimate the maximum luminance values, the error margin was consistently within 5% for all conditions.<i>Conclusion:</i>These results demonstrate the feasibility of implementing an efficient QC system for medical displays and the usefulness of a multitask-based method. Thus, this study provides valuable insights into the potential to reduce the workload associated with medical-device management the development of QC systems for medical devices, highlighting the importance of future efforts to improve their accuracy and applicability.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142943681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-15DOI: 10.1088/2057-1976/ada6bb
Dagbjört Helga Eiríksdóttir, Gry Grønborg Hvass, Henrik Zimmermann, Johannes Jan Struijk, Samuel Emil Schmidt
Fetal phonocardiography is a well-known auscultation technique for evaluation of fetal health. However, murmurs that are synchronous with the maternal heartbeat can often be heard while listening to fetal heart sounds. Maternal placental murmurs (MPM) could be used to detect maternal cardiovascular and placental abnormalities, but the recorded MPMs are often contaminated by ambient interference and noise.Objective:The aim of this study was to compare noise reduction methods to reduce noise in the recorded MPMs. Approach:1) Bandpass filtering (BPF), 2) a multichannel noise reduction (MCh) using either Wiener filter (WF), Least-mean-square or Independent component analysis, 3) a combination of BPF with wavelet transient reduction (WTR) and 4) a combination of MCh and WTR. The methods were tested on signals recorded with two microphone units placed on the abdomen of pregnant women with an electrocardiogram (ECG) recorded simultaneously. The performance was evaluated using coherence and heart cycle duration error (HCDError) as compared with the ECG. Results: The mean of the absolute HCDErrorwas 32.7 ms for the BPF with all methods significantly lower (p < 0.05) than BPF. The lowest errors were obtained for WTR-WF where the HCDErrorranged 16.68-17.72 ms for seven different filter orders. All methods had significantly different coherence measure compared with BPF (p < 0.05). The lowest coherence was reached with WTR-WF (filter order 640) where the mean value decreased from 0.50 for BPF to 0.03.Significance:These results show how noise reduction techniques such as WF combined with wavelet denoising can greatly enhance the quality of MPM recordings.
{"title":"Noise reduction in abdominal acoustic recordings of maternal placental murmurs.","authors":"Dagbjört Helga Eiríksdóttir, Gry Grønborg Hvass, Henrik Zimmermann, Johannes Jan Struijk, Samuel Emil Schmidt","doi":"10.1088/2057-1976/ada6bb","DOIUrl":"10.1088/2057-1976/ada6bb","url":null,"abstract":"<p><p>Fetal phonocardiography is a well-known auscultation technique for evaluation of fetal health. However, murmurs that are synchronous with the maternal heartbeat can often be heard while listening to fetal heart sounds. Maternal placental murmurs (MPM) could be used to detect maternal cardiovascular and placental abnormalities, but the recorded MPMs are often contaminated by ambient interference and noise.<i>Objective:</i>The aim of this study was to compare noise reduction methods to reduce noise in the recorded MPMs<i>. Approach:</i>1) Bandpass filtering (BPF), 2) a multichannel noise reduction (MCh) using either Wiener filter (WF), Least-mean-square or Independent component analysis, 3) a combination of BPF with wavelet transient reduction (WTR) and 4) a combination of MCh and WTR. The methods were tested on signals recorded with two microphone units placed on the abdomen of pregnant women with an electrocardiogram (ECG) recorded simultaneously. The performance was evaluated using coherence and heart cycle duration error (HCD<sub>Error</sub>) as compared with the ECG. R<i>esults</i>: The mean of the absolute HCD<sub>Error</sub>was 32.7 ms for the BPF with all methods significantly lower (p < 0.05) than BPF. The lowest errors were obtained for WTR-WF where the HCD<sub>Error</sub>ranged 16.68-17.72 ms for seven different filter orders. All methods had significantly different coherence measure compared with BPF (p < 0.05). The lowest coherence was reached with WTR-WF (filter order 640) where the mean value decreased from 0.50 for BPF to 0.03.<i>Significance:</i>These results show how noise reduction techniques such as WF combined with wavelet denoising can greatly enhance the quality of MPM recordings.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142943680","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Objective
Applying carbon ion beams, which have high linear energy transfer and low scatter within the human body, to Spatially Fractionated Radiation Therapy (SFRT) could benefit the treatment of deep-seated or radioresistant tumors. This study aims to simulate the dose distributions of spatially fractionated beams (SFB) to accurately determine the delivered dose and model the cell survival rate following SFB irradiation.
Approach
Dose distributions of carbon ion beams are calculated using the Triple Gaussian Model. The sensitive volume of the detector used in measurements was also considered. If the measurements and simulations show good agreement, the dose distribution and absolute dose delivered by SFB can be accurately estimated. Three types of dose distributions were delivered to human salivary gland cells (HSGc-C5): uniform dose distribution (UDD), and one-dimensional (1D) grid-like dose distributions (GDD) with 6 mm and 8 mm spacing. These provided high (Peak-to-Valley Dose Ratio, PVDR=4.0) and low (PVDR=1.64) dose differences between peak and valley doses, respectively. Linear-Quadratic (LQ) model parameters for HSGc-C5 were derived from the UDD and cell survival fractions (SF) were simulated for 1D GDD using these values.
Main results
Good agreement was observed between measurements and simulations when accounting for detector volume. However, the TPS results overestimated dose in steep gradient region, likely due to the 2.0 mm calculation grid size. LQ parameters for HSGc-C5 were α = 0.34 and β = 0.057. The 1D GDD with 6 mm spacing showed good agreement between simulations and experiments, but the 8.0 mm spacing resulted in lower experimental cell survival.
Significance
We successfully simulated grid-like dose distributions and conducted SF simulations. The results suggest potential cell-killing effects due to high-dose differences in SFB.
.
{"title":"Quantitative Assessment of Delivered Dose in Carbon Ion Spatially Fractionated Radiotherapy (C-SFRT) and Biological Response to C-SFRT.","authors":"Toshiro Tsubouchi, Misato Umemura, Kazumasa Minami, Noriaki Hamatani, Naoto Saruwatari, Masaaki Takashina, Masashi Yagi, Keith M Furutani, Shinichi Shimizu, Tatsuaki Kanai","doi":"10.1088/2057-1976/ada964","DOIUrl":"https://doi.org/10.1088/2057-1976/ada964","url":null,"abstract":"<p><p>Objective
Applying carbon ion beams, which have high linear energy transfer and low scatter within the human body, to Spatially Fractionated Radiation Therapy (SFRT) could benefit the treatment of deep-seated or radioresistant tumors. This study aims to simulate the dose distributions of spatially fractionated beams (SFB) to accurately determine the delivered dose and model the cell survival rate following SFB irradiation.
Approach
Dose distributions of carbon ion beams are calculated using the Triple Gaussian Model. The sensitive volume of the detector used in measurements was also considered. If the measurements and simulations show good agreement, the dose distribution and absolute dose delivered by SFB can be accurately estimated. Three types of dose distributions were delivered to human salivary gland cells (HSGc-C5): uniform dose distribution (UDD), and one-dimensional (1D) grid-like dose distributions (GDD) with 6 mm and 8 mm spacing. These provided high (Peak-to-Valley Dose Ratio, PVDR=4.0) and low (PVDR=1.64) dose differences between peak and valley doses, respectively. Linear-Quadratic (LQ) model parameters for HSGc-C5 were derived from the UDD and cell survival fractions (SF) were simulated for 1D GDD using these values.
Main results
Good agreement was observed between measurements and simulations when accounting for detector volume. However, the TPS results overestimated dose in steep gradient region, likely due to the 2.0 mm calculation grid size. LQ parameters for HSGc-C5 were α = 0.34 and β = 0.057. The 1D GDD with 6 mm spacing showed good agreement between simulations and experiments, but the 8.0 mm spacing resulted in lower experimental cell survival.
Significance
We successfully simulated grid-like dose distributions and conducted SF simulations. The results suggest potential cell-killing effects due to high-dose differences in SFB.
.</p>","PeriodicalId":8896,"journal":{"name":"Biomedical Physics & Engineering Express","volume":" ","pages":""},"PeriodicalIF":1.3,"publicationDate":"2025-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142999524","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}