Model Decision Research Articles

A multispatial information representation model emphasizing key brain regions for Alzheimer's disease diagnosis with structural magnetic resonance imaging.

Structural magnetic resonance imaging (sMRI) can reflect structural abnormalities of the brain. Due to its high tissue contrast and spatial resolution, it is considered as an MRI sequence in diagnostic tasks related to Alzheimer's disease (AD). Thus far, most studies based on sMRI have only focused on pathological changes in disease-related brain regions in Euclidean space, ignoring the association and interaction between brain regions represented in non-Euclidean space. This non-Euclidean spatial information can provide valuable information for brain disease research. However, few studies have combined Euclidean spatial information in images and graph spatial information in brain networks for the early diagnosis of AD. The purpose of this study is to explore how to effectively combine multispatial information for enhancing AD diagnostic performance. A multispatial information representation model (MSRNet) was constructed for the diagnosis of AD using sMRI. Specifically, the MSRNet included a Euclidean representation channel integrating a multiscale module and a feature enhancement module, in addition to a graph (non-Euclidean) representation channel integrating a node feature aggregation mechanism. This was accomplished through the adoption of a multilayer graph convolutional neural network and a node connectivity aggregation mechanism with fully connected layers. Each participants' gray-matter volume map and preconstructed radiomics-based morphology brain network (radMBN) were used as MSRNet inputs for the learning of multispatial information. Other than the multispatial information representation in MSRNet, an interactive mechanism was proposed to connect the Euclidean and graph representation channels by five disease-related brain regions which were identified based on a classifier operated on with two feature strategies of voxel intensities and radiomics features. MSRNet focused on disease-related brain regions while integrating multispatial information to effectively enhance disease discrimination. The MSRNet was validated on four publicly available datasets, achieving accuracies 92.8% and 90.6% for AD in intra-database and inter-database cross-validation, respectively. The accuracy of MSRNet in distinguishing between late mild cognitive impairment (MCI) and early MCI, and between progressive MCI and stable MCI, reached 79.8% and 73.4%, respectively. The experiments demonstrated that the model's decision scores exhibited good detection capability for MCI progression. Furthermore, the potential of decision scores for improving diagnostic performance was exhibited by combining decision scores with other clinical indicators for AD identification. The MSRNet model could conduct an effective multispatial information representation in the sMRI-based diagnosis of AD. The proposed interaction mechanism in the MSRNet could help the model focus on AD-related brain regions, thus further improving the diagnostic ability.

CovNet-UFCSPA: auxiliando no diagnóstico de pneumonia por coronavírus

Objective: This study introduces the CovNet-UFCSPA architecture, which incorporates pre-processing data from clinical images (X-rays) and deep learning algorithms. Method: A total of 24,235 images were used for training, validation, and testing of the model, identifying areas in the X-rays that influence the model's decision. Result: The architecture achieved a recall of 99% in classifying X-rays from patients at the Hospital de Clínicas de Porto Alegre (HCPA). The application of the CLAHE technique improved the region of interest in the X-rays, reducing the false negative rate from 187 to 9. Conclusion: Compared with Resnet50 V2 and Inception V3 architectures, CovNet-UFCSPA demonstrated superiority in false negative rates, true positives, and recall.

Ensemble learning based Convolutional Neural Network – Depth Fire for detecting COVID-19 in Chest X-Ray images

The Unique Corona virus-caused COVID19 deadly disease has gave out a significant dispute to healthcare systems around the world. To stop the virus's transmission and lessen its negative effects on public health, it is crucial to recognise correctly and rapidly those who have COVID19. The application of artificial intelligence (AI) holds the capacity to increase the precision and effectiveness of COVID19 diagnosis. The purpose of the study is to build a reliable AI-based model capable correctly detect COVID19 cases from chest X-ray pictures. A dataset of 16,000 chest X-ray pictures, including COVID19 positive and negative instances, is employed in the investigation. Four convolutional neural network (CNN) the models that previously been trained are employed in the proposed model, and the output of each model is combined using an ensembling technique. The major objective of this project is to develop an accurate and reliable AI-based model to classify COVID19 cases from chest X-ray images. The individuality of this method comes in its capacity to employ data augmentation strategies to enhance model generalisation and prevent overfitting. The accuracy and dependability of the model are moreover advanced by utilising numerous pre-trained CNN models and ensembling methods. The suggested AI-based model's classification accuracy for the five classes (bacterial, COVID19 positive, negative, opacity, and viral), the three classes (COVID19 positive, negative, and healthy), and the two classes (COVID19 positive and negative) was 97.3%, 98.2%, 97.6%, and respectively. The projected model performs better in terms of sensitivity, accuracy and specificity than unconventional techniques that are previously in use. Significant ability may be guided in the suggested AI-based model's ability to recognize COVID19 cases quickly and effectively from X-rays of the chest. This approach can help radiology physicians analyse affected role quickly and correctly, improving patient outcomes and lessening the strain on healthcare systems. To ensure the precision of the diagnosis, it is vital to mention that the model's decisions should be made in consultation with a licenced medical expert.

Vulnerability detection using BERT based LLM model with transparency obligation practice towards trustworthy AI

Vulnerabilities in the source code are one of the main causes of potential threats in software-intensive systems. There are a large number of vulnerabilities published each day, and effective vulnerability detection is critical to identifying and mitigating these vulnerabilities. AI has emerged as a promising solution to enhance vulnerability detection, offering the ability to analyse vast amounts of data and identify patterns indicative of potential threats. However, AI-based methods often face several challenges, specifically when dealing with large datasets and understanding the specific context of the problem. Large Language Model (LLM) is now widely considered to tackle more complex tasks and handle large datasets, which also exhibits limitations in terms of explaining the model outcome and existing works focus on providing overview of explainability and transparency. This research introduces a novel transparency obligation practice for vulnerability detection using BERT based LLMs. We address the black-box nature of LLMs by employing XAI techniques, unique combination of SHAP, LIME, heat map. We propose an architecture that combines the BERT model with transparency obligation practices, which ensures the assurance of transparency throughout the entire LLM life cycle. An experiment is performed with a large source code dataset to demonstrate the applicability of the proposed approach. The result shows higher accuracy of 91.8 % for the vulnerability detection and model explainability outcome is highly influenced by “vulnerable”, “function”, "mysql_tmpdir_list", “strmov” tokens using both SHAP and LIME framework. Heatmap of attention weights, highlights the local token interactions that aid in understanding the model's decision points.

DeepCNP: An Efficient White-Box Testing of Deep Neural Networks by Aligning Critical Neuron Paths

ContextErroneous decisions of Deep Neural Networks may pose a significant threat to Deep Learning systems deployed in security-critical domains. The key to testing DNNs is to propose a testing technique to generate test cases that can detect more defects of the models. It has been demonstrated that coverage-guided fuzz testing methods are difficult to detect the correctness defects of model's decision logic. Meanwhile, the neuron activation threshold is set based on experience, which increases the uncertainty of the test even more. In addition, the randomly selected seed mutations are prone to generate a large number of invalid test cases, which has a great impact on the testing efficiency. ObjectiveThis paper introduces DeepCNP, a method that combines Critical Neuron Paths alignment and dynamic seeds selection strategy, which can comprehensively and efficiently test all the decision paths of DNN and generate as many different classes of test cases as possible to expose misbehaviors of the model and thus finding defects. MethodDeepCNP utilizes training data to construct decision paths determined by the neuron output distribution, and aligns different decision paths in order to generate test cases. Seeds that are easy to align are dynamically selected based on the decision paths to be tested, and the labeling of seed mutations is specified during the path alignment process, thus improving the efficiency of fuzz testing. ResultsExperimental results show that DeepCNP achieves new state-of-the-art results, pioneering the testing of all decision logics of the model through critical neuron path alignment, which greatly enhances the number of defects found, the efficiency and number of generated test cases. ConclusionDeepCNP comprehensively tests the decision logic of DNNs, efficiently generating a large number of test cases of different categories to expose model's misbehaviors and thus finding additional defects.

Adversarial Attacks in Explainable Machine Learning: A Survey of Threats Against Models and Humans

ABSTRACTReliable deployment of machine learning models such as neural networks continues to be challenging due to several limitations. Some of the main shortcomings are the lack of interpretability and the lack of robustness against adversarial examples or out‐of‐distribution inputs. In this paper, we comprehensively review the possibilities and limits of adversarial attacks for explainable machine learning models. First, we extend the notion of adversarial examples to fit in explainable machine learning scenarios where a human assesses not only the input and the output classification, but also the explanation of the model's decision. Next, we propose a comprehensive framework to study whether (and how) adversarial examples can be generated for explainable models under human assessment. Based on this framework, we provide a structured review of the diverse attack paradigms existing in this domain, identify current gaps and future research directions, and illustrate the main attack paradigms discussed. Furthermore, our framework considers a wide range of relevant yet often ignored factors such as the type of problem, the user expertise or the objective of the explanations, in order to identify the attack strategies that should be adopted in each scenario to successfully deceive the model (and the human). The intention of these contributions is to serve as a basis for a more rigorous and realistic study of adversarial examples in the field of explainable machine learning.

Early-stage cardiomegaly detection and classification from X-ray images using convolutional neural networks and transfer learning

Cardiomyopathy is a serious condition that can result in heart failure, sudden cardiac death, malignant arrhythmias, and thromboembolism. It is a significant contributor to morbidity and mortality globally. The initial finding of cardiomegaly on radiological imaging may signal a deterioration of a known heart condition, an unknown heart disease, or a heart complication related to another illness. Further cardiological evaluation is needed to confirm the diagnosis and determine appropriate treatment. A chest radiograph (X-ray) is the main imaging method used to identify cardiomegaly when the heart is enlarged. A prompt and accurate diagnosis is essential to help healthcare providers determine the most appropriate treatment options before the condition worsens. This study aims to utilize convolutional neural networks and transfer learning techniques, specifically Inception, DenseNet-169, and ResNet-50, to classify cardiomegaly from chest X-ray images automatically. The utilization of block-matching and 3D filtering (BM3D) techniques aimed at enhancing image edge retention, decreasing noise, and utilizing contrast limited adaptive histogram equalization (CLAHE) to enhance contrast in low-intensity images. Gradient-weighted Class Activation Mapping (GradCAM) was used to visualize the significant activation regions contributing to the model's decision. After evaluating all the models, the ResNet-50 model showed outstanding performance. It achieved perfect accuracy of 100 % in both training, and validation, and an impressive 99.8 % accuracy in testing. Additionally, it displayed complete 100 % precision, recall, and F1-score. These findings demonstrate that ResNet-50 surpassed all other models in the study. As a result, the impressive performance of the ResNet-50 model suggests that it could be a valuable tool in improving the efficiency and accuracy of cardiomyopathy diagnosis, ultimately leading to better patient outcomes.

Development of a transformer-based deep learning algorithm for diabetic peripheral neuropathy classification using corneal confocal microscopy images.

Diabetic peripheral neuropathy (DPN) is common and can go unnoticed until it is firmly developed. This study aims to establish a transformer-based deep learning algorithm (DLA) to classify corneal confocal microscopy (CCM) images, identifying DPN in diabetic patients. Our classification model differs from traditional convolutional neural networks (CNNs) using a Swin transformer network with a hierarchical architecture backbone. Participants included those with (DPN+, n = 57) or without (DPN-, n = 37) DPN as determined by the updated Toronto consensus criteria. The CCM image dataset (consisting of 570 DPN+ and 370 DPN- images, with five images selected from each participant's left and right eyes) was randomly divided into training, validation, and test subsets at a 7:1:2 ratio, considering individual participants. The effectiveness of the algorithm was assessed using diagnostic accuracy measures, such as sensitivity, specificity, and accuracy, in conjunction with Grad-CAM visualization techniques to interpret the model's decisions. In the DPN + group (n = 12), the transformer model successfully predicted all participants, while in the DPN- group (n = 7), one participant was misclassified as DPN+, with an area under the curve (AUC) of 0.9405 (95% CI 0.8166, 1.0000). Among the DPN + images (n = 120), 117 were correctly classified, and among the DPN- images (n = 70), 49 were correctly classified, with an AUC of 0.8996 (95% CI 0.8502, 0.9491). For single-image predictions, the transformer model achieved a superior AUC relative to the ResNet50 model (0.8761, 95% CI 0.8155, 0.9366), the Inception_v3 model (0.8802, 95% CI 0.8231, 0.9374), and the DenseNet121 model (0.8965, 95% CI 0.8438, 0.9491). Transformer-based networks outperform CNN-based networks in rapid binary DPN classification. Transformer-based DLAs have clinical DPN screening potential.

Enhancing the Interpretability of Malaria and Typhoid Diagnosis with Explainable AI and Large Language Models.

Malaria and Typhoid fever are prevalent diseases in tropical regions, and both are exacerbated by unclear protocols, drug resistance, and environmental factors. Prompt and accurate diagnosis is crucial to improve accessibility and reduce mortality rates. Traditional diagnosis methods cannot effectively capture the complexities of these diseases due to the presence of similar symptoms. Although machine learning (ML) models offer accurate predictions, they operate as "black boxes" with non-interpretable decision-making processes, making it challenging for healthcare providers to comprehend how the conclusions are reached. This study employs explainable AI (XAI) models such as Local Interpretable Model-agnostic Explanations (LIME), and Large Language Models (LLMs) like GPT to clarify diagnostic results for healthcare workers, building trust and transparency in medical diagnostics by describing which symptoms had the greatest impact on the model's decisions and providing clear, understandable explanations. The models were implemented on Google Colab and Visual Studio Code because of their rich libraries and extensions. Results showed that the Random Forest model outperformed the other tested models; in addition, important features were identified with the LIME plots while ChatGPT 3.5 had a comparative advantage over other LLMs. The study integrates RF, LIME, and GPT in building a mobile app to enhance the interpretability and transparency in malaria and typhoid diagnosis system. Despite its promising results, the system's performance is constrained by the quality of the dataset. Additionally, while LIME and GPT improve transparency, they may introduce complexities in real-time deployment due to computational demands and the need for internet service to maintain relevance and accuracy. The findings suggest that AI-driven diagnostic systems can significantly enhance healthcare delivery in environments with limited resources, and future works can explore the applicability of this framework to other medical conditions and datasets.

Combined Physics- and Machine-Learning-Based Method to Identify Druggable Binding Sites Using SILCS-Hotspots.

Identifying druggable binding sites on proteins is an important and challenging problem, particularly for cryptic, allosteric binding sites that may not be obvious from X-ray, cryo-EM, or predicted structures. The Site-Identification by Ligand Competitive Saturation (SILCS) method accounts for the flexibility of the target protein using all-atom molecular simulations that include various small molecule solutes in aqueous solution. During the simulations, the combination of protein flexibility and comprehensive sampling of the water and solute spatial distributions can identify buried binding pockets absent in experimentally determined structures. Previously, we reported a method for leveraging the information in the SILCS sampling to identify binding sites (termed Hotspots) of small mono- or bicyclic compounds, a subset of which coincide with known binding sites of drug-like molecules. Here, we build on that physics-based approach and present a ML model for ranking the Hotspots according to the likelihood they can accommodate drug-like molecules (e.g., molecular weight >200 Da). In the independent validation set, which includes various enzymes and receptors, our model recalls 67% and 89% of experimentally validated ligand binding sites in the top 10 and 20 ranked Hotspots, respectively. Furthermore, we show that the model's output Decision Function is a useful metric to predict binding sites and their potential druggability in new targets. Given the utility the SILCS method for ligand discovery and optimization, the tools presented represent an important advancement in the identification of orthosteric and allosteric binding sites and the discovery of drug-like molecules targeting those sites.

Unveiling the Power of Model-Agnostic Multiscale Analysis for Enhancing Artificial Intelligence Models in Breast Cancer Histopathology Images.

Developing AI models for digital pathology has traditionally relied on single-scale analysis of histopathology slides. However, a whole slide image is a rich digital representation of the tissue, captured at various magnification levels. Limiting our analysis to a single scale overlooks critical information, spanning from intricate high-resolution cellular details to broad low-resolution tissue structures. In this study, we propose a model-agnostic multiresolution feature aggregation framework tailored for the analysis of histopathology slides in the context of breast cancer, on a multicohort dataset of 2038 patient samples. We have adapted 9 state-of-the-art multiple instance learning models on our multi-scale methodology and evaluated their performance on grade prediction, TP53 mutation status prediction and survival prediction. The results prove the dominance of the multiresolution methodology, and specifically, concatenating or linearly transforming via a learnable layer the feature vectors of image patches from a high (20x) and low (10x) magnification factors achieve improved performance for all prediction tasks across domain-specific and imagenet-based features. On the contrary, the performance of uniresolution baseline models was not consistent across domain-specific and imagenet-based features. Moreover, we shed light on the inherent inconsistencies observed in models trained on whole-tissue-sections when validated against biopsy-based datasets. Despite these challenges, our findings underscore the superiority of multiresolution analysis over uniresolution methods. Finally, cross-scale analysis also benefits the explainability aspects of attention-based architectures, since one can extract attention maps at the tissue- and cell-levels, improving the interpretation of the model's decision.

A Surrogate Model's Decision Tree Method Evaluation for Uncertainty Quantification on a Finite Element Structure via a Fuzzy-Random Approach

A novel additive manufacturing method (AM)) constructs a three-dimensional model from a computer-aided design by adding material layer by layer. This technique produces a lightweight end product with complex geometries and has gained recognition among industrial players. Nonetheless, the mechanical properties and geometry components are the uncertainties that prevail in its structures. An alternative approach using the Finite Element Method (FEM) to analyse these uncertainties demands extensive computational effort and time consumption. Therefore, a machine learning (ML) tool using the surrogate modelling technique offers an alternative way to provide and predict simulation outcomes. This study applies two surrogate modeling approaches, the decision tree (DT) and the Gaussian process regression (GPR) methods. Output data from a FEM simulation with uncertainty elements are obtained for the training purposes of the surrogate models. Both ML methods can predict simulation results with high precision. Both approaches obtained an excellent coefficient of determination value, R2 of 0.998, and Root Mean Square Error, RMSE of 0.012, successfully reducing time consumption and computational effort. The DT method shows better robustness when compared to the GPR method. A value change in the input parameter significantly impacts the surrogate model's prediction performance. An adequate quantity of data input for the training phase of both surrogate models exhibits the FEM results with the presence of uncertainty and robustness.

Concrete Crack Detection and Segregation: A Feature Fusion, Crack Isolation, and Explainable AI-Based Approach.

Scientific knowledge of image-based crack detection methods is limited in understanding their performance across diverse crack sizes, types, and environmental conditions. Builders and engineers often face difficulties with image resolution, detecting fine cracks, and differentiating between structural and non-structural issues. Enhanced algorithms and analysis techniques are needed for more accurate assessments. Hence, this research aims to generate an intelligent scheme that can recognize the presence of cracks and visualize the percentage of cracks from an image along with an explanation. The proposed method fuses features from concrete surface images through a ResNet-50 convolutional neural network (CNN) and curvelet transform handcrafted (HC) method, optimized by linear discriminant analysis (LDA), and the eXtreme gradient boosting (XGB) classifier then uses these features to recognize cracks. This study evaluates several CNN models, including VGG-16, VGG-19, Inception-V3, and ResNet-50, and various HC techniques, such as wavelet transform, counterlet transform, and curvelet transform for feature extraction. Principal component analysis (PCA) and LDA are assessed for feature optimization. For classification, XGB, random forest (RF), adaptive boosting (AdaBoost), and category boosting (CatBoost) are tested. To isolate and quantify the crack region, this research combines image thresholding, morphological operations, and contour detection with the convex hulls method and forms a novel algorithm. Two explainable AI (XAI) tools, local interpretable model-agnostic explanations (LIMEs) and gradient-weighted class activation mapping++ (Grad-CAM++) are integrated with the proposed method to enhance result clarity. This research introduces a novel feature fusion approach that enhances crack detection accuracy and interpretability. The method demonstrates superior performance by achieving 99.93% and 99.69% accuracy on two existing datasets, outperforming state-of-the-art methods. Additionally, the development of an algorithm for isolating and quantifying crack regions represents a significant advancement in image processing for structural analysis. The proposed approach provides a robust and reliable tool for real-time crack detection and assessment in concrete structures, facilitating timely maintenance and improving structural safety. By offering detailed explanations of the model's decisions, the research addresses the critical need for transparency in AI applications, thus increasing trust and adoption in engineering practice.

SwinDFU-Net: Deep learning transformer network for infection identification in diabetic foot ulcer.

The identification of infection in diabetic foot ulcers (DFUs) is challenging due to variability within classes, visual similarity between classes, reduced contrast with healthy skin, and presence of artifacts. Existing studies focus on visual characteristics and tissue classification rather than infection detection, critical for assessing DFUs and predicting amputation risk. To address these challenges, this study proposes a deep learning model using a hybrid CNN and Swin Transformer architecture for infection classification in DFU images. The aim is to leverage end-to-end mapping without prior knowledge, integrating local and global feature extraction to improve detection accuracy. The proposed model utilizes a hybrid CNN and Swin Transformer architecture. It employs the Grad CAM technique to visualize the decision-making process of the CNN and Transformer blocks. The DFUC Challenge dataset is used for training and evaluation, emphasizing the model's ability to accurately classify DFU images into infected and non-infected categories. The model achieves high performance metrics: sensitivity (95.98%), specificity (97.08%), accuracy (96.52%), and Matthews Correlation Coefficient (0.93). These results indicate the model's effectiveness in quickly diagnosing DFU infections, highlighting its potential as a valuable tool for medical professionals. The hybrid CNN and Swin Transformer architecture effectively combines strengths from both models, enabling accurate classification of DFU images as infected or non-infected, even in complex scenarios. The use of Grad CAM provides insights into the model's decision process, aiding in identifying infected regions within DFU images. This approach shows promise for enhancing clinical assessment and management of DFU infections.

Novel approach exploring the correlation between presepsin and routine laboratory parameters using explainable artificial intelligence

Although presepsin, a crucial biomarker for the diagnosis and management of sepsis, has gained prominence in contemporary medical research, its relationship with routine laboratory parameters, including demographic data and hospital blood test data, remains underexplored. This study integrates machine learning with explainable artificial intelligence (XAI) to provide insights into the relationship between presepsin and these parameters. Advanced machine learning classifiers provide a multilateral view of data and play an important role in highlighting the interrelationships between presepsin and other parameters. XAI enhances analysis by ensuring transparency in the model's decisions, especially in selecting key parameters that significantly enhance classification accuracy. Utilizing XAI, this study successfully identified critical parameters that increased the predictive accuracy for sepsis patients, achieving a remarkable ROC AUC of 0.97 and an accuracy of 0.94. This breakthrough is possibly attributed to the comprehensive utilization of XAI in refining parameter selection, thus leading to these significant predictive metrics. The presence of missing data in datasets is another concern; this study addresses it by employing Extreme Gradient Boosting (XGBoost) to manage missing data, effectively mitigating potential biases while preserving both the accuracy and relevance of the results. The perspective of examining data from higher dimensions using machine learning transcends traditional observation and analysis. The findings of this study hold the potential to enhance patient diagnoses and treatment, underscoring the value of merging traditional research methods with advanced analytical tools.

M-XAF: Medical explainable diagnosis system of atrial fibrillation based on medical knowledge and semantic representation fusion

BackgroundElectrocardiogram (ECG) is a critical diagnostic tool used for screening atrial fibrillation (AF). In recent years, several deep learning-based ECG automatic diagnosis methods have been proposed and have yielded satisfactory results. However, the lack of interpretability in these systems limits their clinical usage. MethodsTo address this issue, we propose a medical explainable AF diagnosis system (M-XAF) that not only predicts AF diagnostic results but also generates an analysis report of the ECG medical features that influenced the classification model's decision. Using a novel ECG signal features occlusion analysis method, M-XAF establishes a connection between the medical knowledge space and the representation space from the convolutional neural network (CNN) by extracting understandable semantic sensitive channels (SSCs) via gradient-weighted channel sensitivity. Through the visualization and statistical modeling of SSCs, M-XAF provides multimodal explanations, including ECG feature descriptions, regions of interest (ROI), and class activation mapping (CAM), for AF diagnostic results. ConclusionThe experimental results demonstrate that the M-XAF has achieved satisfactory results in the AF identification task, with accuracy, specificity, PPV, F1, and MCC reaching over 0.9715, 0.9896, 0.9778, 0.9567, and 0.9365 on two public ECG datasets. In addition, M-XAF generates quantitative explainable reports and CAM for each predicted sample from the perspective of classification models. Furthermore, the explainable reports provided by M-XAF can address the high-confidence problem in classification models. Ultimately, the explainable framework in M-XAF can be widely applied in AI-ECG classification scenarios.

Understanding the Role of Self-Attention in a Transformer Model for the Discrimination of SCD From MCI Using Resting-State EEG.

The identification of EEG biomarkers to discriminate Subjective Cognitive Decline (SCD) from Mild Cognitive Impairment (MCI) conditions is a complex task which requires great clinical effort and expertise. We exploit the self-attention component of the Transformer architecture to obtain physiological explanations of the model's decisions in the discrimination of 56 SCD and 45 MCI patients using resting-state EEG. Specifically, an interpretability workflow leveraging attention scores and time-frequency analysis of EEG epochs through Continuous Wavelet Transform is proposed. In the classification framework, models are trained and validated with 5-fold cross-validation and evaluated on a test set obtained by selecting 20% of the total subjects. Ablation studies and hyperparameter tuning tests are conducted to identify the optimal model configuration. Results show that the best performing model, which achieves acceptable results both on epochs' and patients' classification, is capable of finding specific EEG patterns that highlight changes in the brain activity between the two conditions. We demonstrate the potential of attention weights as tools to guide experts in understanding which disease-relevant EEG features could be discriminative of SCD and MCI.

Cost‐sensitive tree SHAP for explaining cost‐sensitive tree‐based models

AbstractCost‐sensitive ensemble learning as a combination of two approaches, ensemble learning and cost‐sensitive learning, enables generation of cost‐sensitive tree‐based ensemble models using the cost‐sensitive decision tree (CSDT) learning algorithm. In general, tree‐based models characterize nice graphical representation that can explain a model's decision‐making process. However, the depth of the tree and the number of base models in the ensemble can be a limiting factor in comprehending the model's decision for each sample. The CSDT models are widely used in finance (e.g., credit scoring and fraud detection) but lack effective explanation methods. We previously addressed this gap with cost‐sensitive tree Shapley Additive Explanation Method (CSTreeSHAP), a cost‐sensitive tree explanation method for the single‐tree CSDT model. Here, we extend the introduced methodology to cost‐sensitive ensemble models, particularly cost‐sensitive random forest models. The paper details the theoretical foundation and implementation details of CSTreeSHAP for both single CSDT and ensemble models. The usefulness of the proposed method is demonstrated by providing explanations for single and ensemble CSDT models trained on well‐known benchmark credit scoring datasets. Finally, we apply our methodology and analyze the stability of explanations for those models compared to the cost‐insensitive tree‐based models. Our analysis reveals statistically significant differences between SHAP values despite seemingly similar global feature importance plots of the models. This highlights the value of our methodology as a comprehensive tool for explaining CSDT models.

Histopathologic image-based deep learning classifier for predicting platinum-based treatment responses in high-grade serous ovarian cancer.

Platinum-based chemotherapy is the cornerstone treatment for female high-grade serous ovarian carcinoma (HGSOC), but choosing an appropriate treatment for patients hinges on their responsiveness to it. Currently, no available biomarkers can promptly predict responses to platinum-based treatment. Therefore, we developed the Pathologic Risk Classifier for HGSOC (PathoRiCH), a histopathologic image-based classifier. PathoRiCH was trained on an in-house cohort (n = 394) and validated on two independent external cohorts (n = 284 and n = 136). The PathoRiCH-predicted favorable and poor response groups show significantly different platinum-free intervals in all three cohorts. Combining PathoRiCH with molecular biomarkers provides an even more powerful tool for the risk stratification of patients. The decisions of PathoRiCH are explained through visualization and a transcriptomic analysis, which bolster the reliability of our model's decisions. PathoRiCH exhibits better predictive performance than current molecular biomarkers. PathoRiCH will provide a solid foundation for developing an innovative tool to transform the current diagnostic pipeline for HGSOC.

Information fusion-based Bayesian optimized heterogeneous deep ensemble model based on longitudinal neuroimaging data

The fusion of multimodal longitudinal data is difficult but crucial for enhancing the accuracy of deep learning models for disease identification and helps provide tailored and patient-centric decisions. This study explores the fusion of multimodal data to detect the progression of Alzheimer’s disease (AD) using ensemble learning. We propose a heterogeneous ensemble framework of Bayesian-optimized time-series deep learning models to identify progressive deterioration of brain damage. Experimental results show that the heterogeneous ensemble of three models with patient’s temporal data outperforms all other variants of ensemble models by achieving an average performance of 95% for accuracy. We also propose a novel explainability approach, which enables domain experts and practitioners to better comprehend the model's final decision. The visual explainability of infected brain regions and the model's robustness is evaluated by our two medical domain experts showing its promising use in real medical environment. To evaluate the model’s generalizability and robustness, our optimized model is tested on a dataset with different distribution. The experiments demonstrate that the proposed model, which was trained on ADNI data, exhibits reliable generalization to NACC data with an average precision of 90%, recall of 91%, F1-score of 89%, AUC of 88%, and accuracy of 88%.