Eur Respir Rev. 2024 Mar 27;33(171):230055. doi: 10.1183/16000617.0055-2023. Print 2024 Jan 31.

ABSTRACT

The shortcomings of qualitative visual assessment have led to the development of computer-based tools to characterise and quantify disease on high-resolution computed tomography (HRCT) in patients with interstitial lung diseases (ILDs). Quantitative CT (QCT) software enables quantification of patterns on HRCT with results that are objective, reproducible, sensitive to change and predictive of disease progression. Applications developed to provide a diagnosis or pattern classification are mainly based on artificial intelligence. Deep learning, which identifies patterns in high-dimensional data and maps them to segmentations or outcomes, can be used to identify the imaging patterns that most accurately predict disease progression. Optimisation of QCT software will require the implementation of protocol standards to generate data of sufficient quality for use in computerised applications and the identification of diagnostic, imaging and physiological features that are robustly associated with mortality for use as anchors in the development of algorithms. Consortia such as the Open Source Imaging Consortium have a key role to play in the collation of imaging and clinical data that can be used to identify digital imaging biomarkers that inform diagnosis, prognosis and response to therapy.

PMID:38537949 | DOI:10.1183/16000617.0055-2023

Am J Pathol. 2024 Mar 25:S0002-9440(24)00110-X. doi: 10.1016/j.ajpath.2024.02.016. Online ahead of print.

ABSTRACT

Tumor-infiltrating lymphocytes (TILs) are associated with improved survival in patients with epithelial ovarian cancer. However, the evaluation of TILs has not been applied to routine clinical practice due to reproducibility issues. We developed two convolutional neural network models to detect TILs and to determine their spatial location in whole-slide images, and established a spatial assessment pipeline to objectively quantify intraepithelial and stromal TILs in patients with high-grade serous ovarian carcinoma. The predictions of the established models showed a significant positive correlation with the number of CD8+ T cells and immune gene expressions. We demonstrated that patients with a higher density of intraepithelial TILs had a significantly prolonged overall survival (OS) and progression-free survival (PFS) in multiple cohorts. Based on the density of intraepithelial and stromal TILs, we classified patients into three immunophenotypes: immune-inflamed, excluded, and desert. The immune-desert subgroup showed the worst prognosis. Gene expression analysis showed that the immune-desert subgroup had lower immune cytolytic activity (CYT) and T-cell-inflamed gene-expression profile (GEP) scores, whereas immune-excluded subgroup had higher expression of interferon-γ and programmed death 1 receptor (PD-1) signaling pathway. Our established evaluation method provides detailed and comprehensive quantification of intraepithelial and stromal TILs throughout hematoxylin and eosin (H&E)-stained slides, and has potential for clinical application for personalized treatment of patients with ovarian cancer.

PMID:38537936 | DOI:10.1016/j.ajpath.2024.02.016

Neuroimage. 2024 Mar 25:120579. doi: 10.1016/j.neuroimage.2024.120579. Online ahead of print.

ABSTRACT

Very preterm (VPT) infants (born at less than 32 weeks gestational age) are at high risk for various adverse neurodevelopmental deficits. Unfortunately, most of these deficits cannot be accurately diagnosed until the age of 2-5 years old. Given the benefits of early interventions, accurate diagnosis and prediction soon after birth are urgently needed for VPT infants. Previous studies have applied deep learning models to learn the brain structural connectome (SC) to predict neurodevelopmental deficits in the preterm population. However, none of these models are specifically designed for graph-structured data, and thus may potentially miss certain topological information conveyed in the brain SC. In this study, we aim to develop deep learning models to learn the SC acquired at term-equivalent age for early prediction of neurodevelopmental deficits at 2 years corrected age in VPT infants. We directly treated the brain SC as a graph, and applied graph convolutional network (GCN) models to capture complex topological information of the SC. In addition, we applied the supervised contrastive learning (SCL) technique to mitigate the effects of the data scarcity problem, and enable robust training of GCN models. We hypothesize that SCL will enhance GCN models for early prediction of neurodevelopmental deficits in VPT infants using the SC. We used a regional prospective cohort of ∼280 VPT infants who underwent MRI examinations at term-equivalent age from the Cincinnati Infant Neurodevelopment Early Prediction Study (CINEPS). These VPT infants completed neurodevelopmental assessment at 2 years corrected age to evaluate cognition, language, and motor skills. Using the SCL technique, the GCN model achieved mean areas under the receiver operating characteristic curve (AUCs) in the range of 0.72∼0.75 for predicting three neurodevelopmental deficits, outperforming several competing models. Our results support our hypothesis that the SCL technique is able to enhance the GCN model in our prediction tasks.

PMID:38537766 | DOI:10.1016/j.neuroimage.2024.120579

Comput Biol Med. 2024 Mar 23;173:108365. doi: 10.1016/j.compbiomed.2024.108365. Online ahead of print.

ABSTRACT

BACKGROUND: Most of the methods using digital pathological image for predicting Hepatocellular carcinoma (HCC) prognosis have not considered paracancerous tissue microenvironment (PTME), which are potentially important for tumour initiation and metastasis. This study aimed to identify roles of image features of PTME in predicting prognosis and tumour recurrence of HCC patients.

METHODS: We collected whole slide images (WSIs) of 146 HCC patients from Sun Yat-sen Memorial Hospital (SYSM dataset). For each WSI, five types of regions of interests (ROIs) in PTME and tumours were manually annotated. These ROIs were used to construct a Lasso Cox survival model for predicting the prognosis of HCC patients. To make the model broadly useful, we established a deep learning method to automatically segment WSIs, and further used it to construct a prognosis prediction model. This model was tested by the samples of 225 HCC patients from the Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC).

RESULTS: In predicting prognosis of the HCC patients, using the image features of manually annotated ROIs in PTME achieved C-index 0.668 in the SYSM testing dataset, which is higher than the C-index 0.648 reached by the model only using image features of tumours. Integrating ROIs of PTME and tumours achieved C-index 0.693 in the SYSM testing dataset. The model using automatically segmented ROIs of PTME and tumours achieved C-index of 0.665 (95% CI: 0.556-0.774) in the TCGA-LIHC samples, which is better than the widely used methods, WSISA (0.567), DeepGraphSurv (0.593), and SeTranSurv (0.642). Finally, we found the Texture SumAverage Skew HV on immune cell infiltration and Texture related features on desmoplastic reaction are the most important features of PTME in predicting HCC prognosis. We additionally used the model in prediction HCC recurrence for patients from SYSM-training, SYSM-testing, and TCGA-LIHC datasets, indicating the important roles of PTME in the prediction.

CONCLUSIONS: Our results indicate image features of PTME is critical for improving the prognosis prediction of HCC. Moreover, the image features related with immune cell infiltration and desmoplastic reaction of PTME are the most important factors associated with prognosis of HCC.

PMID:38537563 | DOI:10.1016/j.compbiomed.2024.108365

Comput Med Imaging Graph. 2024 Mar 23;114:102376. doi: 10.1016/j.compmedimag.2024.102376. Online ahead of print.

ABSTRACT

Acute ischemic stroke is a critical health condition that requires timely intervention. Following admission, clinicians typically use perfusion imaging to facilitate treatment decision-making. While deep learning models leveraging perfusion data have demonstrated the ability to predict post-treatment tissue infarction for individual patients, predictions are often represented as binary or probabilistic masks that are not straightforward to interpret or easy to obtain. Moreover, these models typically rely on large amounts of subjectively segmented data and non-standard perfusion analysis techniques. To address these challenges, we propose a novel deep learning approach that directly predicts follow-up computed tomography images from full spatio-temporal 4D perfusion scans through a temporal compression. The results show that this method leads to realistic follow-up image predictions containing the infarcted tissue outcomes. The proposed compression method achieves comparable prediction results to using perfusion maps as inputs but without the need for perfusion analysis or arterial input function selection. Additionally, separate models trained on 45 patients treated with thrombolysis and 102 treated with thrombectomy showed that each model correctly captured the different patient-specific treatment effects as shown by image difference maps. The findings of this work clearly highlight the potential of our method to provide interpretable stroke treatment decision support without requiring manual annotations.

PMID:38537536 | DOI:10.1016/j.compmedimag.2024.102376

Med Image Anal. 2024 Mar 22;94:103146. doi: 10.1016/j.media.2024.103146. Online ahead of print.

ABSTRACT

Focused cardiac ultrasound (FoCUS) is a valuable point-of-care method for evaluating cardiovascular structures and function, but its scope is limited by equipment and operator's experience, resulting in primarily qualitative 2D exams. This study presents a novel framework to automatically estimate the 3D spatial relationship between standard FoCUS views. The proposed framework uses a multi-view U-Net-like fully convolutional neural network to regress line-based heatmaps representing the most likely areas of intersection between input images. The lines that best fit the regressed heatmaps are then extracted, and a system of nonlinear equations based on the intersection between view triplets is created and solved to determine the relative 3D pose between all input images. The feasibility and accuracy of the proposed pipeline were validated using a novel realistic in silico FoCUS dataset, demonstrating promising results. Interestingly, as shown in preliminary experiments, the estimation of the 2D images' relative poses enables the application of 3D image analysis methods and paves the way for 3D quantitative assessments in FoCUS examinations.

PMID:38537416 | DOI:10.1016/j.media.2024.103146

Int J Med Robot. 2024 Feb;20(1):e2620. doi: 10.1002/rcs.2620.

ABSTRACT

BACKGROUND: Swift and accurate decision-making is pivotal in managing intestinal obstructions. This study aims to integrate deep learning and surgical expertise to enhance decision-making in intestinal obstruction cases.

METHODS: We developed a deep learning model based on the YOLOv8 framework, trained on a dataset of 700 images categorised into operated and non-operated groups, with surgical outcomes as ground truth. The model's performance was evaluated through standard metrics.

RESULTS: At a confidence threshold of 0.5, the model demonstrated sensitivity of 83.33%, specificity of 78.26%, precision of 81.7%, recall of 75.1%, and mAP@0.5 of 0.831.

CONCLUSIONS: The model exhibited promising outcomes in distinguishing operative and nonoperative management cases. The fusion of deep learning with surgical expertise enriches decision-making in intestinal obstruction management. The proposed model can assist surgeons in intricate scenarios such as intestinal obstruction management and promotes the synergy between technology and clinical acumen for advancing patient care.

PMID:38536723 | DOI:10.1002/rcs.2620

IEEE Trans Neural Netw Learn Syst. 2024 Mar 27;PP. doi: 10.1109/TNNLS.2024.3379172. Online ahead of print.

ABSTRACT

Few-shot image classification (FSIC) is beneficial for a variety of real-world scenarios, aiming to construct a recognition system with limited training data. In this article, we extend the original FSIC task by incorporating defense against malicious adversarial examples. This can be an arduous challenge because numerous deep learning-based approaches remain susceptible to adversarial examples, even when trained with ample amounts of data. Previous studies on this problem have predominantly concentrated on the meta-learning framework, which involves sampling numerous few-shot tasks during the training stage. In contrast, we propose a straightforward but effective baseline via learning robust and discriminative representations without tedious meta-task sampling, which can further be generalized to unforeseen adversarial FSIC tasks. Specifically, we introduce an adversarial-aware (AA) mechanism that exploits feature-level distinctions between the legitimate and the adversarial domains to provide supplementary supervision. Moreover, we design a novel adversarial reweighting training strategy to ameliorate the imbalance among adversarial examples. To further enhance the adversarial robustness without compromising discriminative features, we propose the cyclic feature purifier during the postprocessing projection, which can reduce the interference of unforeseen adversarial examples. Furthermore, our method can obtain robust feature embeddings that maintain superior transferability, even when facing cross-domain adversarial examples. Extensive experiments and systematic analyses demonstrate that our method achieves state-of-the-art robustness as well as natural performance among adversarially robust FSIC algorithms on three standard benchmarks by a substantial margin.

PMID:38536695 | DOI:10.1109/TNNLS.2024.3379172

IEEE Trans Haptics. 2024 Mar 27;PP. doi: 10.1109/TOH.2024.3382258. Online ahead of print.

ABSTRACT

Current Virtual Reality (VR) environments lack the haptic signals that humans experience during real-life interactions, such as the sensation of texture during lateral movement on a surface. Adding realistic haptic textures to VR environments requires a model that generalizes to variations of a user's interaction and to the wide variety of existing textures in the world. Current methodologies for haptic texture rendering exist, but they usually develop one model per texture, resulting in low scalability. We present a deep learning-based action-conditional model for haptic texture rendering and evaluate its perceptual performance in rendering realistic texture vibrations through a multi-part human user study. This model is unified over all materials and uses data from a vision-based tactile sensor (GelSight) to render the appropriate surface conditioned on the user's action in real-time. For rendering texture, we use a high-bandwidth vibrotactile transducer attached to a 3D Systems Touch device. The results of our user study shows that our learning-based method creates high-frequency texture renderings with comparable or better quality than state-of-the-art methods without the need to learn a separate model per texture. Furthermore, we show that the method is capable of rendering previously unseen textures using a single GelSight image of their surface.

PMID:38536688 | DOI:10.1109/TOH.2024.3382258

IEEE J Biomed Health Inform. 2024 Mar 27;PP. doi: 10.1109/JBHI.2024.3382604. Online ahead of print.

ABSTRACT

Deep learning in ultrasound(US) imaging aims to construct foundational models that accurately reflect the modality's unique characteristics. Nevertheless, the limited datasets and narrow task types have restricted this field in recent years. To address these challenges, we introduce US-MTD120K, a multi-task ultrasound dataset with 120,354 real-world two-dimensional images. This dataset covers three standard plane recognition and two diagnostic tasks in ultrasound imaging, providing a rich basis for model training and evaluation. We detail the data collection, distribution, and labelling processes, ensuring a thorough understanding of the dataset's structure. Furthermore, we conduct extensive benchmark tests on 27 state-of-the-art methods from both supervised and self-supervised learning(SSL) perspectives. In the realm of supervised learning, we analyze the sensitivity of two main feature computation methods to ultrasound images at the representational level, highlighting that models which judiciously constrain global feature computation could potentially serve as a viable analytical approach for US image analysis. In the context of self-supervised learning, we delved into the modelling process of self-supervised learning models for medical images and proposed an improvement strategy, named MoCo-US, a solution that addresses the excessive reliance on pretext task design from the input side. It achieves competitive performance with minimal pretext task design and enhances other SSL methods simply. The dataset and the code will be available at https://github.com/JsongZhang/CDOA-for-UMTD.

PMID:38536687 | DOI:10.1109/JBHI.2024.3382604

IEEE J Biomed Health Inform. 2024 Mar 27;PP. doi: 10.1109/JBHI.2024.3381896. Online ahead of print.

ABSTRACT

Molecular property prediction is an important task in drug discovery. However, experimental data for many drug molecules are limited, especially for novel molecular structures or rare diseases which affect the accuracy of many deep learning methods that rely on large training datasets. To this end, we propose PG-DERN, a novel few-shot learning model for molecular property prediction. A dual-view encoder is introduced to learn a meaningful molecular representation by integrating information from node and subgraph. Next, a relation graph learning module is proposed to construct a relation graph based on the similarity between molecules, which improves the efficiency of information propagation and the accuracy of property prediction. In addition, we use a MAML-based meta-learning strategy to learn well-initialized meta-parameters. In order to guide the tuning of meta-parameters, a property-guided feature augmentation module is designed to transfer information from similar properties to the novel property to improve the comprehensiveness of the feature representation of molecules with novel property. A series of comparative experiments on four benchmark datasets demonstrate that the proposed PG-DERN outperforms state-of-the-art methods.

PMID:38536684 | DOI:10.1109/JBHI.2024.3381896

IEEE/ACM Trans Comput Biol Bioinform. 2024 Mar 27;PP. doi: 10.1109/TCBB.2024.3381825. Online ahead of print.

ABSTRACT

Protein-to-protein interaction (PPI) prediction aims to predict whether two given proteins interact or not. Compared with traditional experimental methods of high cost and low efficiency, the current deep learning based approach makes it possible to discover massive potential PPIs from large-scale databases. However, deep PPI prediction models perform poorly on unseen species, as their proteins are not in the training set. Targetting on this issue, the paper first proposes PPITrans, a Transformer based PPI prediction model that exploits a language model pre-trained on proteins to conduct binary PPI prediction. To validate the effectiveness on unseen species, PPITrans is trained with Human PPIs and tested on PPIs of other species. Experimental results show that PPITrans significantly outperforms the previous state-of-the-art on various metrics, especially on PPIs of unseen species. For example, the AUPR improves 0.339 absolutely on Fly PPIs. Aiming to explore the knowledge learned by PPITrans from PPI data, this paper also designs a series of probes belonging to three categories. Their results reveal several interesting findings, like that although PPITrans cannot capture the spatial structure of proteins, it can obtain knowledge of PPI type and binding affinity, learning more than binary PPI.

PMID:38536676 | DOI:10.1109/TCBB.2024.3381825

Insights Imaging. 2024 Mar 27;15(1):99. doi: 10.1186/s13244-024-01661-4.

ABSTRACT

OBJECTIVES: This retrospective single-center analysis aimed to evaluate whether artificial intelligence can detect type 2 diabetes mellitus by evaluating the pectoral muscle on digital breast tomosynthesis (DBT).

MATERIAL METHOD: An analysis of 11,594 DBT images of 287 consecutive female patients (mean age 60, range 40-77 years) was conducted using convolutional neural networks (EfficientNetB5). The inclusion criterion was left-sided screening images with unsuspicious interpretation who also had a current glycosylated hemoglobin A1c (HBA1c) % value. The exclusion criteria were inadequate imaging, history of breast cancer, and/or diabetes mellitus. HbA1c values between 5.6 and 6.4% were categorized as prediabetic, and those with values ≥ 6.5% were categorized as diabetic. A recorded HbA1c ≤ 5.5% served as the control group. Each group was divided into 3 subgroups according to age. Images were subjected to pattern analysis parameters then cropped and resized in a format to contain only pectoral muscle. The dataset was split into 85% for training and 15% for testing the model's performance. The accuracy rate and F1-score were selected as performance indicators.

RESULTS: The training process was concluded in the 15th epoch, each comprising 1000 steps, with an accuracy rate of 92% and a loss of only 0.22. The average specificity and sensitivity for all 3 groups were 95%. The F1-score was 0.95. AUC-ROC was 0.995. PPV was 94%, and NPV was 98%.

CONCLUSION: Our study presented a pioneering approach, applying deep learning for the detection of diabetes mellitus status in women using pectoral muscle images and was found to function with an accuracy rate of 92%.

CRITICAL RELEVANCE STATEMENT: AI can differentiate pathological changes within pectoral muscle tissue by assessing radiological images and maybe a potential diagnostic tool for detecting diabetes mellitus and other diseases that affect muscle tissues.

KEY POINTS: • AI may have an opportunistic use as a screening exam for diabetes during digital breast tomosynthesis. • This technique allows for early and non-invasive detection of diabetes mellitus by AI. • AI may have broad applications in detecting pathological changes within muscle tissue.

PMID:38536577 | DOI:10.1186/s13244-024-01661-4

Jpn J Radiol. 2024 Mar 27. doi: 10.1007/s11604-024-01550-2. Online ahead of print.

ABSTRACT

PURPOSE: Radiotherapy planning incorporating functional lung images has the potential to reduce pulmonary toxicity. Free-breathing 4DCT-derived ventilation image (CTVI) may help quantify lung function. This study introduces a novel deep-learning model directly translating planning CT images into CTVI. We investigated the accuracy of generated images and the impact on functional avoidance planning.

MATERIALS AND METHODS: Paired planning CT and 4DCT scans from 48 patients with NSCLC were randomized to training (n = 41) and testing (n = 7) data sets. The ventilation maps were generated from 4DCT using a Jacobian-based algorithm to provide ground truth labels (CTVI4DCT). A 3D U-Net-based model was trained to map CT to synthetic CTVI (CTVISyn) and validated using fivefold cross-validation. The highest-performing model was applied to the testing set. Spearman's correlation (rs) and Dice similarity coefficient (DSC) determined voxel-wise and functional-wise concordance between CTVI4DCT and CTVISyn. Three plans were designed per patient in the testing set: one clinical plan without CTVI and two functional avoidance plans combined with CTVI4DCT or CTVISyn, aimed at sparing high-functional lungs defined as the top 50% of the percentile ventilation ranges. Dose-volume (DVH) parameters regarding the planning target volume (PTV) and organs at risk (OARs) were recorded. Radiation pneumonitis (RP) risk was estimated using a dose-function (DFH)-based normal tissue complication probability (NTCP) model.

RESULTS: CTVISyn showed a mean rs value of 0.65 ± 0.04 compared to CTVI4DCT. Mean DSC values over the top 50% and 60% of ventilation ranges were 0.41 ± 0.07 and 0.52 ± 0.10, respectively. In the test set (n = 7), all patients' RP-risk benefited from CTVI4DCT-guided plans (Riskmean_4DCT_vs_Clinical: 29.24% vs. 49.12%, P = 0.016), and six patients benefited from CTVISyn-guided plans (Riskmean_Syn_vs_Clinical: 31.13% vs. 49.12%, P = 0.022). There were no significant differences in DVH and DFH metrics between CTVISyn and CTVI4DCT-guided plan (P > 0.05).

CONCLUSION: Using deep-learning techniques, CTVISyn generated from planning CT exhibited a moderate-to-high correlation with CTVI4DCT. The CTVISyn-guided plans were comparable to the CTVI4DCT-guided plans, effectively reducing pulmonary toxicity in patients while maintaining acceptable plan quality. Further prospective trials are needed to validate these findings.

PMID:38536558 | DOI:10.1007/s11604-024-01550-2

Eur Radiol. 2024 Mar 27. doi: 10.1007/s00330-024-10714-7. Online ahead of print.

ABSTRACT

OBJECTIVE: To investigate the effect of uncertainty estimation on the performance of a Deep Learning (DL) algorithm for estimating malignancy risk of pulmonary nodules.

METHODS AND MATERIALS: In this retrospective study, we integrated an uncertainty estimation method into a previously developed DL algorithm for nodule malignancy risk estimation. Uncertainty thresholds were developed using CT data from the Danish Lung Cancer Screening Trial (DLCST), containing 883 nodules (65 malignant) collected between 2004 and 2010. We used thresholds on the 90th and 95th percentiles of the uncertainty score distribution to categorize nodules into certain and uncertain groups. External validation was performed on clinical CT data from a tertiary academic center containing 374 nodules (207 malignant) collected between 2004 and 2012. DL performance was measured using area under the ROC curve (AUC) for the full set of nodules, for the certain cases and for the uncertain cases. Additionally, nodule characteristics were compared to identify trends for inducing uncertainty.

RESULTS: The DL algorithm performed significantly worse in the uncertain group compared to the certain group of DLCST (AUC 0.62 (95% CI: 0.49, 0.76) vs 0.93 (95% CI: 0.88, 0.97); p < .001) and the clinical dataset (AUC 0.62 (95% CI: 0.50, 0.73) vs 0.90 (95% CI: 0.86, 0.94); p < .001). The uncertain group included larger benign nodules as well as more part-solid and non-solid nodules than the certain group.

CONCLUSION: The integrated uncertainty estimation showed excellent performance for identifying uncertain cases in which the DL-based nodule malignancy risk estimation algorithm had significantly worse performance.

CLINICAL RELEVANCE STATEMENT: Deep Learning algorithms often lack the ability to gauge and communicate uncertainty. For safe clinical implementation, uncertainty estimation is of pivotal importance to identify cases where the deep learning algorithm harbors doubt in its prediction.

KEY POINTS: • Deep learning (DL) algorithms often lack uncertainty estimation, which potentially reduce the risk of errors and improve safety during clinical adoption of the DL algorithm. • Uncertainty estimation identifies pulmonary nodules in which the discriminative performance of the DL algorithm is significantly worse. • Uncertainty estimation can further enhance the benefits of the DL algorithm and improve its safety and trustworthiness.

PMID:38536463 | DOI:10.1007/s00330-024-10714-7

Cancer Res. 2024 Mar 27. doi: 10.1158/0008-5472.CAN-23-0860. Online ahead of print.

ABSTRACT

T cells recognize tumor antigens and initiate an anti-cancer immune response in the very early stages of tumor development, and the antigen specificity of T cells is determined by the T cell receptor (TCR). Therefore, monitoring changes in the TCR repertoire in peripheral blood may offer a strategy to detect various cancers at a relatively early stages. Here, we developed the deep learning framework iCanTCR to identify cancer patients based on the TCR repertoire. The iCanTCR framework uses TCRβ sequences from an individual as an input and outputs the predicted cancer probability. The model was trained on over 2000 publicly available TCR repertoires from eleven types of cancer and healthy controls. Analysis of several additional publicly available datasets validated the ability of iCanTCR to distinguish cancer patients from non-cancer individuals and demonstrated the capability of iCanTCR for the accurate classification of multiple cancers. Importantly, iCanTCR precisely identified individuals with early-stage cancer with an area under the curve (AUC) of 86%. Altogether, this work provides a liquid biopsy approach to capture immune signals from peripheral blood for non-invasive cancer diagnosis.

PMID:38536129 | DOI:10.1158/0008-5472.CAN-23-0860

Med Phys. 2024 Mar 27. doi: 10.1002/mp.17038. Online ahead of print.

ABSTRACT

BACKGROUND: Stereotactic body radiotherapy of thoracic and abdominal tumors has to account for respiratory intrafractional tumor motion. Commonly, an external breathing signal is continuously acquired that serves as a surrogate of the tumor motion and forms the basis of strategies like breathing-guided imaging and gated dose delivery. However, due to inherent system latencies, there exists a temporal lag between the acquired respiratory signal and the system response. Respiratory signal prediction models aim to compensate for the time delays and to improve imaging and dose delivery.

PURPOSE: The present study explores and compares six state-of-the-art machine and deep learning-based prediction models, focusing on real-time and real-world applicability. All models and data are provided as open source and data to ensure reproducibility of the results and foster reuse.

METHODS: The study was based on 2502 breathing signals ( t t o t a l ≈ 90 $t_{total} \approx 90$ h) acquired during clinical routine, split into independent training (50%), validation (20%), and test sets (30%). Input signal values were sampled from noisy signals, and the target signal values were selected from corresponding denoised signals. A standard linear prediction model (Linear), two state-of-the-art models in general univariate signal prediction (Dlinear, Xgboost), and three deep learning models (Lstm, Trans-Enc, Trans-TSF) were chosen. The prediction performance was evaluated for three different prediction horizons (480, 680, and 920 ms). Moreover, the robustness of the different models when applied to atypical, that is, out-of-distribution (OOD) signals, was analyzed.

RESULTS: The Lstm model achieved the lowest normalized root mean square error for all prediction horizons. The prediction errors only slightly increased for longer horizons. However, a substantial spread of the error values across the test signals was observed. Compared to typical, that is, in-distribution test signals, the prediction accuracy of all models decreased when applied to OOD signals. The more complex deep learning models Lstm and Trans-Enc showed the least performance loss, while the performance of simpler models like Linear dropped the most. Except for Trans-Enc, inference times for the different models allowed for real-time application.

CONCLUSION: The application of the Lstm model achieved the lowest prediction errors. Simpler prediction filters suffer from limited signal history access, resulting in a drop in performance for OOD signals.

PMID:38536107 | DOI:10.1002/mp.17038

Nanomaterials (Basel). 2024 Mar 15;14(6):527. doi: 10.3390/nano14060527.

ABSTRACT

Neuromorphic computing has emerged as an alternative computing paradigm to address the increasing computing needs for data-intensive applications. In this context, resistive random access memory (RRAM) devices have garnered immense interest among the neuromorphic research community due to their capability to emulate intricate neuronal behaviors. RRAM devices excel in terms of their compact size, fast switching capabilities, high ON/OFF ratio, and low energy consumption, among other advantages. This review focuses on the multifaceted aspects of RRAM devices and their application to brain-inspired computing. The review begins with a brief overview of the essential biological concepts that inspire the development of bio-mimetic computing architectures. It then discusses the various types of resistive switching behaviors observed in RRAM devices and the detailed physical mechanisms underlying their operation. Next, a comprehensive discussion on the diverse material choices adapted in recent literature has been carried out, with special emphasis on the benchmark results from recent research literature. Further, the review provides a holistic analysis of the emerging trends in neuromorphic applications, highlighting the state-of-the-art results utilizing RRAM devices. Commercial chip-level applications are given special emphasis in identifying some of the salient research results. Finally, the current challenges and future outlook of RRAM-based devices for neuromorphic research have been summarized. Thus, this review provides valuable understanding along with critical insights and up-to-date information on the latest findings from the field of resistive switching devices towards brain-inspired computing.

PMID:38535676 | DOI:10.3390/nano14060527

J Imaging. 2024 Mar 21;10(3):72. doi: 10.3390/jimaging10030072.

ABSTRACT

We investigate the impact of different data modalities for cattle weight estimation. For this purpose, we collect and present our own cattle dataset representing the data modalities: RGB, depth, combined RGB and depth, segmentation, and combined segmentation and depth information. We explore a recent vision-transformer-based zero-shot model proposed by Meta AI Research for producing the segmentation data modality and for extracting the cattle-only region from the images. For experimental analysis, we consider three baseline deep learning models. The objective is to assess how the integration of diverse data sources influences the accuracy and robustness of the deep learning models considering four different performance metrics: mean absolute error (MAE), root mean squared error (RMSE), mean absolute percentage error (MAPE), and R-squared (R2). We explore the synergies and challenges associated with each modality and their combined use in enhancing the precision of cattle weight prediction. Through comprehensive experimentation and evaluation, we aim to provide insights into the effectiveness of different data modalities in improving the performance of established deep learning models, facilitating informed decision-making for precision livestock management systems.

PMID:38535152 | DOI:10.3390/jimaging10030072

J Imaging. 2024 Mar 15;10(3):71. doi: 10.3390/jimaging10030071.

ABSTRACT

The application of large field-of-view (FoV) cameras equipped with fish-eye lenses brings notable advantages to various real-world computer vision applications, including autonomous driving. While deep learning has proven successful in conventional computer vision applications using regular perspective images, its potential in fish-eye camera contexts remains largely unexplored due to limited datasets for fully supervised learning. Semi-supervised learning comes as a potential solution to manage this challenge. In this study, we explore and benchmark two popular semi-supervised methods from the perspective image domain for fish-eye image segmentation. We further introduce FishSegSSL, a novel fish-eye image segmentation framework featuring three semi-supervised components: pseudo-label filtering, dynamic confidence thresholding, and robust strong augmentation. Evaluation on the WoodScape dataset, collected from vehicle-mounted fish-eye cameras, demonstrates that our proposed method enhances the model's performance by up to 10.49% over fully supervised methods using the same amount of labeled data. Our method also improves the existing image segmentation methods by 2.34%. To the best of our knowledge, this is the first work on semi-supervised semantic segmentation on fish-eye images. Additionally, we conduct a comprehensive ablation study and sensitivity analysis to showcase the efficacy of each proposed method in this research.

PMID:38535151 | DOI:10.3390/jimaging10030071