J Cheminform. 2024 Dec 20;16(1):142. doi: 10.1186/s13321-024-00930-0.

ABSTRACT

Protein-protein interactions (PPIs) play a crucial role in numerous biochemical and biological processes. Although several structure-based molecular generative models have been developed, PPI interfaces and compounds targeting PPIs exhibit distinct physicochemical properties compared to traditional binding pockets and small-molecule drugs. As a result, generating compounds that effectively target PPIs, particularly by considering PPI complexes or interface hotspot residues, remains a significant challenge. In this work, we constructed a comprehensive dataset of PPI interfaces with active and inactive compound pairs. Based on this, we propose a novel molecular generative framework tailored to PPI interfaces, named GENiPPI. Our evaluation demonstrates that GENiPPI captures the implicit relationships between the PPI interfaces and the active molecules, and can generate novel compounds that target these interfaces. Moreover, GENiPPI can generate structurally diverse novel compounds with limited PPI interface modulators. To the best of our knowledge, this is the first exploration of a structure-based molecular generative model focused on PPI interfaces, which could facilitate the design of PPI modulators. The PPI interface-based molecular generative model enriches the existing landscape of structure-based (pocket/interface) molecular generative model. SCIENTIFIC CONTRIBUTION: This study introduces GENiPPI, a protein-protein interaction (PPI) interface-aware molecular generative framework. The framework first employs Graph Attention Networks to capture atomic-level interaction features at the protein complex interface. Subsequently, Convolutional Neural Networks extract compound representations in voxel and electron density spaces. These features are integrated into a Conditional Wasserstein Generative Adversarial Network, which trains the model to generate compound representations targeting PPI interfaces. GENiPPI effectively captures the relationship between PPI interfaces and active/inactive compounds. Furthermore, in fewshot molecular generation, GENiPPI successfully generates compounds comparable to known disruptors. GENiPPI provides an efficient tool for structure-based design of PPI modulators.

PMID:39707457 | DOI:10.1186/s13321-024-00930-0

BMC Med Imaging. 2024 Dec 20;24(1):345. doi: 10.1186/s12880-024-01528-6.

ABSTRACT

INTRODUCTION: Gadolinium-based T1-weighted MRI sequence is the gold standard for the detection of active multiple sclerosis (MS) lesions. The performance of machine learning (ML) and deep learning (DL) models in the classification of active and non-active MS lesions from the T2-weighted MRI images has been investigated in this study.

METHODS: 107 Features of 75 active and 100 non-active MS lesions were extracted by using SegmentEditor and Radiomics modules of 3D slicer software. Sixteen ML and one sequential DL models were created using the 5-fold cross-validation method and each model with its special optimized parameters trained using the training-validation datasets. Models' performances in test data set were evaluated by metric parameters of accuracy, precision, sensitivity, specificity, AUC, and F1 score.

RESULTS: The sequential DL model achieved the highest AUC of 95.60% on the test dataset, demonstrating its superior ability to distinguish between active and non-active plaques. Among traditional ML models, the Hybrid Gradient Boosting Classifier (HGBC) demonstrated a commendable test AUC of 86.75%, while the Gradient Boosting Classifier (GBC) excelled in cross-validation with an AUC of 87.92%.

CONCLUSION: The performance of sixteen ML and one sequential DL models in the classification of active and non-active MS lesions was evaluated. The results of the study highlight the effectiveness of sequential DL approach and ensemble methods in achieving robust predictive performance, underscoring their potential applications in classifying MS plaques.

PMID:39707207 | DOI:10.1186/s12880-024-01528-6

J Med Chem. 2024 Dec 20. doi: 10.1021/acs.jmedchem.4c02668. Online ahead of print.

ABSTRACT

Fluorine (F) substitution is a common method of drug discovery and development. However, there are no accurate approaches available for predicting the bioactivity changes after F-substitution, as the effect of substitution on the interactions between compounds and proteins (CPI) remains a mystery. In this study, we constructed a data set with 111,168 pairs of fluorine-substituted and nonfluorine-substituted compounds. We developed a multimodal deep learning model (F-CPI). In comparison with traditional machine learning and popular CPI task models, the accuracy, precision, and recall of F-CPI (∼90, ∼79, and ∼45%) were higher than those of GraphDTA (∼86, ∼58, and ∼40%). The application of the F-CPI for the structural optimization of hit compounds against SARS-CoV-2 3CLpro by F-substitution achieved a more than 100-fold increase in bioactivity (IC50: 0.23 μM vs 28.19 μM). Therefore, the multimodal deep learning model F-CPI would be a veritable and effective tool in the context of drug discovery and design.

PMID:39707149 | DOI:10.1021/acs.jmedchem.4c02668

J Imaging Inform Med. 2024 Dec 20. doi: 10.1007/s10278-024-01362-w. Online ahead of print.

ABSTRACT

To develop and validate a modality-invariant Swin U-Net Transformer (UNETR) deep learning model for liver and spleen segmentation on abdominal T1-weighted (T1w) or T2-weighted (T2w) MR images from multiple institutions for pediatric and adult patients with known or suspected chronic liver diseases. In this IRB-approved retrospective study, clinical abdominal axial T1w and T2w MR images from pediatric and adult patients were retrieved from four study sites, including Cincinnati Children's Hospital Medical Center (CCHMC), New York University (NYU), University of Wisconsin (UW) and University of Michigan / Michigan Medicine (UM). The whole liver and spleen were manually delineated as the ground truth masks. We developed a modality-invariant 3D Swin UNETR using a modality-invariant training strategy, in which each patient's T1w and T2w MR images were treated as separate training samples. We conducted both internal and external validation experiments. A total of 241 T1w and 339 T2w MR sequences from 304 patients (age [mean ± standard deviation], 31.8 ± 20.3 years; 132 [43%] female) were included for model development. The Swin UNETR achieved a Dice similarity coefficient (DSC) of 0.95 ± 0.02 on T1w images and 0.93 ± 0.05 on T2w images for liver segmentation. This is significantly better than the U-Net model (0.90 ± 0.05, p < 0.001 and 0.90 ± 0.13, p < 0.001, respectively). The Swin UNETR achieved a DSC of 0.88 ± 0.12 on T1w images and 0.93 ± 0.10 on T2w images for spleen segmentation, and it significantly outperformed a modality-invariant U-Net model (0.80 ± 0.18, p = 0.001 and 0.88 ± 0.12, p = 0.002, respectively). Our study demonstrated that a modality-invariant Swin UNETR model can segment the liver and spleen on routinely collected clinical bi-parametric abdominal MR images from children and adult patients.

PMID:39707114 | DOI:10.1007/s10278-024-01362-w

J Imaging Inform Med. 2024 Dec 20. doi: 10.1007/s10278-024-01369-3. Online ahead of print.

ABSTRACT

The accurate and early detection of vertebral metastases is crucial for improving patient outcomes. Although deep-learning models have shown potential in this area, their lack of prediction reliability and robustness limits their clinical utility. To address these challenges, we propose a novel technique called Ensemble Monte Carlo Dropout (EMCD) for uncertainty quantification (UQ), which combines the Monte Carlo dropout and deep ensembles. In this retrospective study, we analyzed 11,468 abdominal computed tomography images from 116 patients diagnosed with vertebral metastases and 957 images from 11 healthy controls. Uncertainty was quantified and visualized using single number, predictive probability interval, posterior distribution and uncertainty class activation maps to provide a detailed understanding of prediction confidence. The EMCD model demonstrated superior performance compared with traditional UQ methods, achieving an area under the receiver operating characteristic curve (AUC) of 0.93 and an expected calibration error of 0.09, indicating high predictive accuracy and reliability. In addition, the model exhibited strong performance in handling out-of-distribution data. When data retention was applied based on uncertainty values, the AUC of the model improved to 0.96, highlighting the potential of uncertainty-driven data selection to enhance performance. The EMCD model represents a significant advancement in the automated detection of vertebral metastases, providing superior diagnostic accuracy and introducing a robust UQ framework to aid clinicians in making informed decisions.

PMID:39707112 | DOI:10.1007/s10278-024-01369-3

Med Biol Eng Comput. 2024 Dec 21. doi: 10.1007/s11517-024-03265-y. Online ahead of print.

ABSTRACT

The correct analysis of medical images requires the medical knowledge and expertise of radiologists to understand, clarify, and explain complex patterns and diagnose diseases. After analyzing, radiologists write detailed and well-structured reports that contribute to the precise and timely diagnosis of patients. However, manually writing reports is often expensive and time-consuming, and it is difficult for radiologists to analyze medical images, particularly images with multiple views and perceptions. It is challenging to accurately diagnose diseases, and many methods are proposed to help radiologists, both traditional and deep learning-based. Automatic report generation is widely used to tackle this issue as it streamlines the process and lessens the burden of manual labeling of images. This paper introduces a systematic literature review with a focus on analyses and evaluating existing research on medical report generation. This SLR follows a proper protocol for the planning, reviewing, and reporting of the results. This review recognizes that the most commonly used deep learning models are encoder-decoder frameworks (45 articles), which provide an accuracy of around 92-95%. Transformers-based models (20 articles) are the second most established method and achieve an accuracy of around 91%. The remaining articles explored in this SLR are attention mechanisms (10), RNN-LSTM (10), Large language models (LLM-10), and graph-based methods (4) with promising results. However, these methods also face certain limitations such as overfitting, risk of bias, and high data dependency that impact their performance. The review not only highlights the strengths and challenges of these methods but also suggests ways to handle them in the future to increase the accuracy and timely generation of medical reports. The goal of this review is to direct radiologists toward methods that lessen their workload and provide precise medical diagnoses.

PMID:39707049 | DOI:10.1007/s11517-024-03265-y

Environ Monit Assess. 2024 Dec 21;197(1):85. doi: 10.1007/s10661-024-13493-2.

ABSTRACT

The urban setting notwithstanding, rice cultivation prevails on the outskirts of Hanoi, with the burning of rice straw in the fields posing a significant challenge. Therefore, it is crucial to conduct spatial mapping of rice distribution, assess dry biomass, and determine emissions from rice straw burning within Hanoi city. The efficacy of the deep convolutional neural networks (DCNN) model has been evident in accurately mapping the spatial distribution of rice in Hanoi, where rice cultivation extensively thrives in suburban areas. In the tropical climate of Vietnam, data derived from synthetic aperture radar (SAR) could serve as a valuable resource for mapping rice fields. Additionally, the amalgamated model, Ant Colony Optimization-eXtreme Gradient Boosting (ACO-XGBoost), could serve as a potent instrument in gauging the aboveground biomass (AGB) of rice within this urban center. The current research reveals the spatial distribution of rice biomass in Hanoi city. Among the six levels of the rice biomass distribution map, the majority of regions in Hanoi city were dominated by the fifth tier, ranging between 3.0 and 4.0 kg/m2. This emerges as a pivotal source of emissions impacting the atmospheric quality of the city. It should be emphasized that the incidence of rice straw burning remains substantial, exceeding 80% in the monitored districts of Hanoi city, notably higher in proximity to the city center. These findings serve a significant function for management and policy making to generate data and calculate air pollution levels in Hanoi.

PMID:39707001 | DOI:10.1007/s10661-024-13493-2

Eur J Pediatr. 2024 Dec 21;184(1):98. doi: 10.1007/s00431-024-05925-5.

ABSTRACT

Atopic dermatitis, food allergy, allergic rhinitis, and asthma are among the most common diseases in childhood. They are heterogeneous diseases, can co-exist in their development, and manifest complex associations with other disorders and environmental and hereditary factors. Elucidating these intricacies by identifying clinically distinguishable groups and actionable risk factors will allow for better understanding of the diseases, which will enhance clinical management and benefit society and affected individuals and families. Artificial intelligence (AI) is a promising tool in this context, enabling discovery of meaningful patterns in complex data. Numerous studies within pediatric allergy have and continue to use AI, primarily to characterize disease endotypes/phenotypes and to develop models to predict future disease outcomes. However, most implementations have used relatively simplistic data from one source, such as questionnaires. In addition, methodological approaches and reporting are lacking. This review provides a practical hands-on guide for conducting AI-based studies in pediatric allergy, including (1) an introduction to essential AI concepts and techniques, (2) a blueprint for structuring analysis pipelines (from selection of variables to interpretation of results), and (3) an overview of common pitfalls and remedies. Furthermore, the state-of-the art in the implementation of AI in pediatric allergy research, as well as implications and future perspectives are discussed.

CONCLUSION: AI-based solutions will undoubtedly transform pediatric allergy research, as showcased by promising findings and innovative technical solutions, but to fully harness the potential, methodologically robust implementation of more advanced techniques on richer data will be needed.

WHAT IS KNOWN: • Pediatric allergies are heterogeneous and common, inflicting substantial morbidity and societal costs. • The field of artificial intelligence is undergoing rapid development, with increasing implementation in various fields of medicine and research.

WHAT IS NEW: • Promising applications of AI in pediatric allergy have been reported, but implementation largely lags behind other fields, particularly in regard to use of advanced algorithms and non-tabular data. Furthermore, lacking reporting on computational approaches hampers evidence synthesis and critical appraisal. • Multi-center collaborations with multi-omics and rich unstructured data as well as utilization of deep learning algorithms are lacking and will likely provide the most impactful discoveries.

PMID:39706990 | DOI:10.1007/s00431-024-05925-5

Commun Med (Lond). 2024 Dec 20;4(1):276. doi: 10.1038/s43856-024-00695-5.

ABSTRACT

BACKGROUND: Molecular profiling of estrogen receptor (ER), progesterone receptor (PR), and ERBB2 (also known as Her2) is essential for breast cancer diagnosis and treatment planning. Nevertheless, current methods rely on the qualitative interpretation of immunohistochemistry and fluorescence in situ hybridization (FISH), which can be costly, time-consuming, and inconsistent. Here we explore the clinical utility of predicting receptor status from digitized hematoxylin and eosin-stained (H&E) slides using machine learning trained and evaluated on a multi-institutional dataset.

METHODS: We developed a deep learning system to predict ER, PR, and ERBB2 statuses from digitized H&E slides and evaluated its utility in three clinical applications: identifying hormone receptor-positive patients, serving as a second-read tool for quality assurance, and addressing intratumor heterogeneity. For development and validation, we collected 19,845 slides from 7,950 patients across six independent cohorts representative of diverse clinical settings.

RESULTS: Here we show that the system identifies 30.5% of patients as hormone receptor-positive, achieving a specificity of 0.9982 and a positive predictive value of 0.9992, demonstrating its ability to determine eligibility for hormone therapy without immunohistochemistry. By restaining and reassessing samples flagged as potential false negatives, we discover 31 cases of misdiagnosed ER, PR, and ERBB2 statuses.

CONCLUSIONS: These findings demonstrate the utility of the system in diverse clinical settings and its potential to improve breast cancer diagnosis. Given the substantial focus of current guidelines on reducing false negative diagnoses, this study supports the integration of H&E-based machine learning tools into workflows for quality assurance.

PMID:39706861 | DOI:10.1038/s43856-024-00695-5

Int J Biol Macromol. 2024 Dec 18:138880. doi: 10.1016/j.ijbiomac.2024.138880. Online ahead of print.

ABSTRACT

Anticancer peptides (ACPs) demonstrate significant potential in clinical cancer treatment due to their ability to selectively target and kill cancer cells. In recent years, numerous artificial intelligence (AI) algorithms have been developed. However, many predictive methods lack sufficient wet lab validation, thereby constraining the progress of models and impeding the discovery of novel ACPs. This study proposes a comprehensive research strategy by introducing CNBT-ACPred, an ACP prediction model based on a three-channel deep learning architecture, supported by extensive in vitro and in vivo experiments. CNBT-ACPred achieved an accuracy of 0.9554 and a Matthews Correlation Coefficient (MCC) of 0.8602. Compared to existing excellent models, CNBT-ACPred increased accuracy by at least 5 % and improved MCC by 15 %. Predictions were conducted on over 3.8 million sequences from Uniprot, along with 100,000 sequences generated by a deep generative model, ultimately identifying 37 out of 41 candidate peptides from >30 species that exhibited effective in vitro tumor inhibitory activity. Among these, tPep14 demonstrated significant anticancer effects in two mouse xenograft models without detectable toxicity. Finally, the study revealed correlations between the amino acid composition, structure, and function of the identified ACP candidates.

PMID:39706427 | DOI:10.1016/j.ijbiomac.2024.138880

Biomed Phys Eng Express. 2024 Dec 20. doi: 10.1088/2057-1976/ada1da. Online ahead of print.

ABSTRACT

Auto-encoders have demonstrated outstanding performance in computer vision tasks such as biomedical imaging, including classification, segmentation, and denoising. Many of the current techniques for image denoising in biomedical applications involve training an autoencoder or convolutional neural network (CNN) using pairs of clean and noisy images. However, these approaches are not realistic because the autoencoder or CNN is trained on known noise and does not generalize well to new noisy distributions. This paper proposes a novel approach for biomedical image denoising using a variational network based on a Bayesian model and deep learning.&#xD;Method: In this study, we aim to denoise biomedical images using a Bayesian approach. In our dataset, each image exhibited a same noise distribution. To achieve this, we first estimate the noise distribution based on Bayesian probability by calculating the posterior distributions, and then proceed with denoising. A loss function that combines the Bayesian prior and autoencoder objectives is used to train the variational network. The proposed method was tested on CT-Scan biomedical image datasets and compared with state-of-the-art denoising techniques.&#xD;Results: The experimental results demonstrate that our method outperforms the existing methods in terms of denoising accuracy, visual quality, and computational efficiency. For instance, we obtained a PSNR of 39.18 dB and an SSIM of 0.9941 with noise intensity std = 10. Our approach can potentially improve the accuracy and reliability of biomedical image analysis, which can have significant implications for clinical diagnosis and treatment planning.&#xD;Conclusion: The proposed method combines the advantages of both Bayesian modeling and variational network to effectively denoise biomedical images.&#xD.

PMID:39705726 | DOI:10.1088/2057-1976/ada1da

Biomed Phys Eng Express. 2024 Dec 20. doi: 10.1088/2057-1976/ada1db. Online ahead of print.

ABSTRACT

Recent studies on graph representation learning in brain tumor learning tasks have garnered significant interest by encoding and learning inherent relationships among the geometric features of tumors. There are serious class imbalance problems that occur on brain tumor MRI datasets. Impressive deep learning models like CNN- and Transformer-based can easily address this problem through their complex model architectures with large parameters.&#xD;However, graph-based networks are not suitable for this approach because of chronic over-smoothing and oscillation convergence problems. To address these challenges at once, we propose novel graph spectral convolutional networks called HeatGSNs, which incorporate eigenfilters and learnable low-pass graph heat kernels to capture geometric similarities within tumor classes. They operate to a continuous feature propagation mechanism derived by the forward finite difference of graph heat kernels, which is approximated by the cosine form for the shift-scaled Chebyshev polynomial and modified Bessel functions, leading to fast and accurate performance achievement. Our experimental results show a best average Dice score of 90%, an average Hausdorff Distance (95%) of 5.45mm, and an average accuracy of 80.11% in the BRATS2021 dataset. Moreover, HeatGSNs require significantly fewer parameters, averaging 1.79M, compared to other existing methods, demonstrating efficiency and effectiveness.

PMID:39705725 | DOI:10.1088/2057-1976/ada1db

Brief Bioinform. 2024 Nov 22;26(1):bbae700. doi: 10.1093/bib/bbae700.

NO ABSTRACT

PMID:39705709 | DOI:10.1093/bib/bbae700

J Med Internet Res. 2024 Dec 20;26:e66648. doi: 10.2196/66648.

ABSTRACT

BACKGROUND: As health care continues to evolve with technological advancements, the integration of artificial intelligence into clinical practices has shown promising potential to enhance patient care and operational efficiency. Among the forefront of these innovations are large language models (LLMs), a subset of artificial intelligence designed to understand, generate, and interact with human language at an unprecedented scale.

OBJECTIVE: This systematic review describes the role of LLMs in improving diagnostic accuracy, automating documentation, and advancing specialist education and patient engagement within the field of gastroenterology and gastrointestinal endoscopy.

METHODS: Core databases including MEDLINE through PubMed, Embase, and Cochrane Central registry were searched using keywords related to LLMs (from inception to April 2024). Studies were included if they satisfied the following criteria: (1) any type of studies that investigated the potential role of LLMs in the field of gastrointestinal endoscopy or gastroenterology, (2) studies published in English, and (3) studies in full-text format. The exclusion criteria were as follows: (1) studies that did not report the potential role of LLMs in the field of gastrointestinal endoscopy or gastroenterology, (2) case reports and review papers, (3) ineligible research objects (eg, animals or basic research), and (4) insufficient data regarding the potential role of LLMs. Risk of Bias in Non-Randomized Studies-of Interventions was used to evaluate the quality of the identified studies.

RESULTS: Overall, 21 studies on the potential role of LLMs in gastrointestinal disorders were included in the systematic review, and narrative synthesis was done because of heterogeneity in the specified aims and methodology in each included study. The overall risk of bias was low in 5 studies and moderate in 16 studies. The ability of LLMs to spread general medical information, offer advice for consultations, generate procedure reports automatically, or draw conclusions about the presumptive diagnosis of complex medical illnesses was demonstrated by the systematic review. Despite promising benefits, such as increased efficiency and improved patient outcomes, challenges related to data privacy, accuracy, and interdisciplinary collaboration remain.

CONCLUSIONS: We highlight the importance of navigating these challenges to fully leverage LLMs in transforming gastrointestinal endoscopy practices.

TRIAL REGISTRATION: PROSPERO 581772; https://www.crd.york.ac.uk/prospero/.

PMID:39705703 | DOI:10.2196/66648

J Neurosurg Spine. 2024 Dec 20:1-10. doi: 10.3171/2024.8.SPINE24722. Online ahead of print.

ABSTRACT

OBJECTIVE: The prognostic significance of body composition phenotypes for survival in patients undergoing surgical intervention for spinal metastases has not yet been elucidated. This study aimed to elucidate the impact of body composition phenotypes on surgical outcomes and 5-year survival.

METHODS: The records of patients treated surgically for spinal metastases between 2010 and 2020 were retrospectively evaluated. A deep learning pipeline assessed preoperative CT scans obtained within 3 months of surgery and identified muscle and fat content and composition. These data were used to categorize patients into 4 body composition phenotypic groups: 1) not sarcopenic, not obese; 2) sarcopenia alone; 3) obesity alone; and 4) sarcopenic obesity (SO). The groups were matched using a comprehensive propensity-matching procedure. Rates of postoperative outcomes and survival were evaluated. Cox proportional hazard models were used to evaluate the influence of body composition phenotypes on 5-year survival. Kaplan-Meier plots were used to evaluate survival probability further.

RESULTS: Following a propensity-matching procedure, 102 matched patient records were identified (not sarcopenic, not obese, n = 24; sarcopenia alone, n = 27; obesity alone, n = 37; and SO, n = 14). SO was found to be associated with a significantly increased mortality risk within 60 months (HR 3.27, 95% CI 1.43-7.48). Kaplan-Meier plots demonstrate evident divergence in survival probability within 5 years among patients in the SO group compared to the others (log-rank test, p = 0.022). Additionally, time to death was also lower in patients with SO (p = 0.018). Significant differences in postoperative ambulation rates were noted among patients with SO (p = 0.048), whereas no preoperative difference existed (p = 0.12). No significant differences in postoperative disposition, length of hospital stay, wound-related complications, or inpatient medical complications were otherwise noted (p > 0.05).

CONCLUSIONS: This study identifies SO as a distinct prognostic factor for increased mortality risk in patients undergoing surgery for spinal metastases, highlighting the complex interplay between body composition and patient outcomes. These findings advocate for integrating body composition analysis into preoperative assessment and tailored postoperative care strategies, promoting personalized treatment plans to improve survival and quality of life for this vulnerable patient population.

PMID:39705691 | DOI:10.3171/2024.8.SPINE24722

JCO Clin Cancer Inform. 2024 Dec;8:e2400107. doi: 10.1200/CCI.24.00107. Epub 2024 Dec 20.

ABSTRACT

PURPOSE: Breast cancer relapses are rarely collected by cancer registries because of logistical and financial constraints. Hence, we investigated natural language processing (NLP), enhanced with state-of-the-art deep learning transformer tools and large language models, to automate relapse identification in the text of computed tomography (CT) reports.

METHODS: We analyzed follow-up CT reports from patients diagnosed with breast cancer between January 1, 2005, and December 31, 2014. The reports were curated and annotated for the presence or absence of local, regional, and distant breast cancer relapses. We performed 10-fold cross-validation to evaluate models identifying different types of relapses in CT reports. Model performance was assessed with classification metrics, reported with 95% confidence intervals.

RESULTS: In our data set of 1,445 CT reports, 799 (55.3%) described any relapse, 72 (5.0%) local relapses, 97 (6.7%) regional relapses, and 743 (51.4%) distant relapses. The any-relapse model achieved an accuracy of 89.6% (87.8-91.1), with a sensitivity of 93.2% (91.4-94.9) and a specificity of 84.2% (80.9-87.1). The local relapse model achieved an accuracy of 94.6% (93.3-95.7), a sensitivity of 44.4% (32.8-56.3), and a specificity of 97.2% (96.2-98.0). The regional relapse model showed an accuracy of 93.6% (92.3-94.9), a sensitivity of 70.1% (60.0-79.1), and a specificity of 95.3% (94.2-96.5). Finally, the distant relapse model demonstrated an accuracy of 88.1% (86.2-89.7), a sensitivity of 91.8% (89.9-93.8), and a specificity of 83.7% (80.5-86.4).

CONCLUSION: We developed NLP models to identify local, regional, and distant breast cancer relapses from CT reports. Automating the identification of breast cancer relapses can enhance data collection about patient outcomes.

PMID:39705642 | DOI:10.1200/CCI.24.00107

Curr Med Res Opin. 2024 Dec 20:1-31. doi: 10.1080/03007995.2024.2445142. Online ahead of print.

ABSTRACT

Breast cancer is a significant health challenge, with accurate and timely diagnosis being critical to effective treatment. Immunohistochemistry (IHC) staining is a widely used technique for the evaluation of breast cancer markers, but manual scoring is time-consuming and can be subject to variability. With the rise of Artificial Intelligence (AI), there is an increasing interest in using machine learning and deep learning approaches to automate the scoring of ER, PR and HER2 biomarker in IHC-stained images for effective treatment. In this narrative literature review, we focus on AI-based techniques for the automated scoring of breast cancer markers in IHC-stained images, specifically Allred, Histochemical (H-Score), and HER2 scoring. We aim to identify the current state-of-the-art approaches, challenges, and potential future research prospect for this area of study. By conducting a comprehensive review of the existing literature, we aim to contribute to the ultimate goal of improving the accuracy and efficiency of breast cancer diagnosis and treatment.

PMID:39705612 | DOI:10.1080/03007995.2024.2445142

Medicine (Baltimore). 2024 Dec 20;103(51):e40723. doi: 10.1097/MD.0000000000040723.

ABSTRACT

PURPOSE: To create a nomogram for accurate prognosis of patients with clear cell renal cell carcinoma (ccRCC) based on computed tomography images.

METHODS: Eight hundred twenty-two ccRCC patients with contrast-enhanced computed tomography images involved in this study were collected. A rectangular region of interest surrounding the tumor was used to extract quantitative radiomics and deep-learning features, which were filtered by Cox proportional hazard regression model and least absolute shrinkage and selection operator. Then the selected features formed a fusion signature, which was assessed by Cox proportional hazard regression model method, Kaplan-Meier analysis, receiver operating characteristic curves, and concordance index (C-index) in different clinical subgroups. Finally, a nomogram constructed with this signature and clinicopathologic risk factors was assessed by C-index and survival calibration curves.

RESULTS: The fusion signature performed better than the radiomics signature. Then we combined this signature and 2 clinicopathologic risk factors. This nomogram showed an increase of about 20% in C-index values when compared to clinical nomogram in both datasets. Its prediction probability was also in good agreement with the actual ratio.

CONCLUSION: The proposed fusion nomogram provided a noninvasive and easy-to-use model for survival prognosis of ccRCC patients in future clinical use, without the requirement to perform a detailed segmentation for radiologists.

PMID:39705434 | DOI:10.1097/MD.0000000000040723

Sci Adv. 2024 Dec 20;10(51):eabg0264. doi: 10.1126/sciadv.abg0264. Epub 2024 Dec 20.

ABSTRACT

Deep learning algorithms can extract meaningful diagnostic features from biomedical images, promising improved patient care in digital pathology. Vision Transformer (ViT) models capture long-range spatial relationships and offer robust prediction power and better interpretability for image classification tasks than convolutional neural network models. However, limited annotated biomedical imaging datasets can cause ViT models to overfit, leading to false predictions due to random noise. To address this, we introduce Training Attention and Validation Attention Consistency (TAVAC), a metric for evaluating ViT model overfitting and quantifying interpretation reproducibility. By comparing high-attention regions between training and testing, we tested TAVAC on four public image classification datasets and two independent breast cancer histological image datasets. Overfitted models showed significantly lower TAVAC scores. TAVAC also distinguishes off-target from on-target attentions and measures interpretation generalization at a fine-grained cellular level. Beyond diagnostics, TAVAC enhances interpretative reproducibility in basic research, revealing critical spatial patterns and cellular structures of biomedical and other general nonbiomedical images.

PMID:39705362 | DOI:10.1126/sciadv.abg0264

PLoS One. 2024 Dec 20;19(12):e0315477. doi: 10.1371/journal.pone.0315477. eCollection 2024.

ABSTRACT

Antimicrobial peptides (AMPs) are excellent at fighting many different infections. This demonstrates how important it is to make new AMPs that are even better at eliminating infections. The fundamental transformation in a variety of scientific disciplines, which led to the emergence of machine learning techniques, has presented significant opportunities for the development of antimicrobial peptides. Machine learning and deep learning are used to predict antimicrobial peptide efficacy in the study. The main purpose is to overcome traditional experimental method constraints. Gram-negative bacterium Escherichia coli is the model organism in this study. The investigation assesses 1,360 peptide sequences that exhibit anti- E. coli activity. These peptides' minimal inhibitory concentrations have been observed to be correlated with a set of 34 physicochemical characteristics. Two distinct methodologies are implemented. The initial method involves utilizing the pre-computed physicochemical attributes of peptides as the fundamental input data for a machine-learning classification approach. In the second method, these fundamental peptide features are converted into signal images, which are then transmitted to a deep learning neural network. The first and second methods have accuracy of 74% and 92.9%, respectively. The proposed methods were developed to target a single microorganism (gram negative E.coli), however, they offered a framework that could potentially be adapted for other types of antimicrobial, antiviral, and anticancer peptides with further validation. Furthermore, they have the potential to result in significant time and cost reductions, as well as the development of innovative AMP-based treatments. This research contributes to the advancement of deep learning-based AMP drug discovery methodologies by generating potent peptides for drug development and application. This discovery has significant implications for the processing of biological data and the computation of pharmacology.

PMID:39705302 | DOI:10.1371/journal.pone.0315477