Classification Of Lesions Research Articles

Deep learning-based MVIT-MLKA model for accurate classification of pancreatic lesions: a multicenter retrospective cohort study.

Accurate differentiation between benign and malignant pancreatic lesions is critical for effective patient management. This study aimed to develop and validate a novel deep learning network using baseline computed tomography (CT) images to predict the classification of pancreatic lesions. This retrospective study included 864 patients (422 men, 442 women) with confirmed histopathological results across three medical centers, forming a training cohort, internal testing cohort, and external validation cohort. A novel hybrid model, Multi-Scale Large Kernel Attention with Mobile Vision Transformer (MVIT-MLKA), was developed, integrating CNN and Transformer architectures to classify pancreatic lesions. The model's performance was compared with traditional machine learning methods and advanced deep learning models. We also evaluated the diagnostic accuracy of radiologists with and without the assistance of the optimal model. Model performance was assessed through discrimination, calibration, and clinical applicability. The MVIT-MLKA model demonstrated superior performance in classifying pancreatic lesions, achieving an AUC of 0.974 (95% CI 0.967-0.980) in the training set, 0.935 (95% CI 0.915-0.954) in the internal testing set, and 0.924 (95% CI 0.902-0.945) in the external validation set, outperforming traditional models and other deep learning models (P < 0.05). Radiologists aided by the MVIT-MLKA model showed significant improvements in diagnostic accuracy and sensitivity compared to those without model assistance (P < 0.05). Grad-CAM visualization enhanced model interpretability by effectively highlighting key lesion areas. The MVIT-MLKA model efficiently differentiates between benign and malignant pancreatic lesions, surpassing traditional methods and significantly improving radiologists' diagnostic performance. The integration of this advanced deep learning model into clinical practice has the potential to reduce diagnostic errors and optimize treatment strategies.

Bilinear Interpolation Augmented Deep Feature Extraction with an Improved Remora Optimization-Based Deep Convolutional Neural Network for Skin Lesion Classification

Bilinear Interpolation Augmented Deep Feature Extraction with an Improved Remora Optimization-Based Deep Convolutional Neural Network for Skin Lesion Classification

Hierarchical skin lesion image classification with prototypical decision tree

Traditional disease classification models often disregard the clinical significance of misclassifications and lack interpretability. To overcome these challenges, we propose a hierarchical prototypical decision tree (HPDT) for skin lesion classification. HPDT combines prototypical networks and decision trees, leveraging a class hierarchy to guide interpretable predictions from general to specific categories. By incorporating a hierarchy-based distance matrix, the model prioritizes less severe misclassifications while maintaining diagnostic accuracy. Evaluated on a dataset of 235,268 dermoscopic images across 65 conditions, HPDT outperforms flat classifiers and existing hierarchical methods in accuracy, error severity reduction, and interpretability. It also generalizes effectively to unseen classes. These results highlight the value of integrating clinical hierarchies into model design and training to improve diagnostic reliability and decision transparency, demonstrating HPDT’s potential for clinical decision support.

Skin Lesion Classification Through Test Time Augmentation and Explainable Artificial Intelligence.

Despite significant advancements in the automatic classification of skin lesions using artificial intelligence (AI) algorithms, skepticism among physicians persists. This reluctance is primarily due to the lack of transparency and explainability inherent in these models, which hinders their widespread acceptance in clinical settings. The primary objective of this study is to develop a highly accurate AI-based algorithm for skin lesion classification that also provides visual explanations to foster trust and confidence in these novel diagnostic tools. By improving transparency, the study seeks to contribute to earlier and more reliable diagnoses. Additionally, the research investigates the impact of Test Time Augmentation (TTA) on the performance of six Convolutional Neural Network (CNN) architectures, which include models from the EfficientNet, ResNet (Residual Network), and ResNeXt (an enhanced variant of ResNet) families. To improve the interpretability of the models' decision-making processes, techniques such as t-distributed Stochastic Neighbor Embedding (t-SNE) and Gradient-weighted Class Activation Mapping (Grad-CAM) are employed. t-SNE is utilized to visualize the high-dimensional latent features of the CNNs in a two-dimensional space, providing insights into how the models group different skin lesion classes. Grad-CAM is used to generate heatmaps that highlight the regions of input images that influence the model's predictions. Our findings reveal that Test Time Augmentation enhances the balanced multi-class accuracy of CNN models by up to 0.3%, achieving a balanced accuracy rate of 97.58% on the International Skin Imaging Collaboration (ISIC 2019) dataset. This performance is comparable to, or marginally better than, more complex approaches such as Vision Transformers (ViTs), demonstrating the efficacy of our methodology.

Association of patellofemoral osteoarthritis on patient-reported outcomes after medial unicompartmental knee arthroplasty: a retrospective cohort study.

In contemporary medial unicompartmental knee arthroplasty (mUKA), non-lateral patellofemoral osteoarthritis (PFOA) is not considered a contraindication. However, we still lack knowledge on the association of PFOA severity on patient reported outcome measures (PROMs) after mUKA. We aimed to examine the association between PFOA severity and PROM-score changes after mUKA. We included 549 mobile-bearing mUKAs. PFOA was graded intraoperatively as 0 = normal cartilage, 1-2 = superficial changes or < 50% of depth, and 3-4 = changes of > 50% of depth or to the bone, using the International Cartilage Repair Society (ICRS) cartilage lesion classification system. All patients completed the Oxford Knee Score (OKS), Activity and Participation Questionnaire (APQ), and Forgotten Joint Score (FJS), preoperatively and 3, 12, and 24 months postoperatively. PROM changes were compared using linear regression models adjusted for sex, age, body mass index, and preoperative PROM score. We found no significant differences in OKS, FJS, and APQ change when comparing group 3-4 with group 0 at any follow-up. When comparing group 1-2 with 0 we found a statistical but not clinical significantly higher change in OKS scores at 24-month follow-up (2.5, 95% confidence interval [CI] 0.36-4.6) and in APQ scores at 24-month follow-up (10.6, CI 1.2-20.0) in favor of group 1-2. Severe PFOA, excluding severe lateral facet PFOA, had no negative association on PROM score development following mobile-bearing mUKA.

In vivo Predictors of Focal Type In-stent Restenosis: A Clinical, Angiographical and Optical Coherence Tomography Study.

Introduction：Few studies have evaluated different patterns of in-stent restenosis by optical coherence tomography (OCT). This study aims to identify in vivo predictors for focal restenosis in patients with in-stent restenosis (ISR). Methods: The study recruited patients with ISR who underwent OCT examination in the Cardiology Department of the Affiliated Hospital of Zunyi Medical University from October 2018 to December 2022. Based on the angiographic classification of ISR lesions, the patients were divided into two groups: the focal group (n=58) and the non-focal group (n=158). Results： The white blood cell count was higher in the non-focal group than those in focal type (7.8±3.0 vs. 6.6±2.1, P = 0.007). The prevalence of lipid rich plaque was higher in patients with focal ISR (65.5% vs. 42.4%, P = 0.003). The occurrence of red thrombus (27.8% vs. 12.1%, P = 0.016) and white thrombus (41.1% vs. 24.1%, P = 0.021) was higher in the non-focal group. Multivariate analysis showed that low density lipoprotein cholesterol C (odds ratio [OR]:3.341, 95% confidence interval [CI]: 1.714-9.784, P = 0.046) was independently associated with focal restenosis. While white blood cell count (OR: 0.814, 95% CI: 0.657-0.913, P = 0.047) and stent malapposition (OR: 0.228, 95% CI: 0.057-0.896, P = 0.037) were independently associated with non-focal restenosis. Conclusion：There were significant differences in clinical baselines and OCT identified morphological characteristics in patients between focal and non-focal group. Low density lipoprotein cholesterol C was independent associated with focal restenosis. White blood cell count and stent malapposition were correlated with non-focal restenosis.

Characterization of hepatocellular carcinomawith CT with deep learning reconstruction compared with iterative reconstruction and 3-Tesla MRI.

This study compared the characteristics of lesions suspicious for hepatocellular carcinoma (HCC) and their LI-RADS classifications in adaptive statistical iterative reconstruction (ASIR) and deep learning reconstruction (DLR) to those of MR images, along with radiologist confidence. This prospective single-center trial included patients who underwent four-phase liver CT and multiphasic contrast-enhanced MRI within 7 days from February to August 2023. The lesion characteristics, LI-RADS classifications and confidence scores according to two radiologists on the ASIR, DLR and MRI techniques were compared. If the patient had at least one lesion, he was included in the HCC group, otherwise in the non-HCC group. MRI being the technique with the best sensitivity, concordance of lesions characteristics and LI-RADS classifications were calculated by weighted kappa between the ASIR and MRI and between the DLR and MRI. The confidence scores are expressed as the means and standard deviations. Eighty-nine patients were enrolled, 52 in the HCC group (67 years ± 9 [mean ± SD], 46 men) and 37 in the non-HCC group (68 years ± 9, 33 men). The concordance coefficient between the LI-RADS classification by ASIR and MRI was 0.64 [0.52; 0.76], showing good agreement, that by DLR and MRI was 0.83 [0.73; 0.92], showing excellent agreement. The diagnostic confidence in ASIR was 3.31 ± 0.95 (mean ± SD) and 3.0 ± 1.11, that in the DLR was 3.9 ± 0.88 and 4.11 ± 0.75, that in the MRI was 4.46 ± 0.80 and 4.57 ± 0.80. DLR provided excellent LI-RADS classification concordance with MRI, whereas ASIR provided good concordance. The radiologists' confidence was greater in the DLR than in the ASIR but remained highest in the MR group. Question Does the use of deep learning reconstructions (DLR) improve LI-RADS classification of suspicious hepatocellular carcinoma lesions compared to adaptive statistical iterative reconstructions (ASIR)? Findings DLR demonstrated superior concordance of LI-RADS classification with MRI compared to ASIR. It also provided greater diagnostic confidence than ASIR. Clinical relevance The use of DLR enhances radiologists' ability to visualize and characterize lesions suspected of being HCC, as well as their LI-RADS classification. Moreover, it also boosts their confidence in interpreting these images.

SL-R-CNN-HHO: multi-class skin lesion classification using region-based convolutional neural networks and harris hawk optimization on the HAM dataset

SL-R-CNN-HHO: multi-class skin lesion classification using region-based convolutional neural networks and harris hawk optimization on the HAM dataset

Lesion classification and diabetic retinopathy grading by integrating softmax and pooling operators into vision transformer.

Diabetic retinopathy grading plays a vital role in the diagnosis and treatment of patients. In practice, this task mainly relies on manual inspection using human visual system. However, the human visual system-based screening process is labor-intensive, time-consuming, and error-prone. Therefore, plenty of automated screening technique have been developed to address this task. Among these techniques, the deep learning models have demonstrated promising outcomes in various types of machine vision tasks. However, most of the medical image analysis-oriented deep learning approaches are built upon the convolutional operations, which might neglect the global dependencies between long-range pixels in the medical images. Therefore, the vision transformer models, which can unveil the associations between global pixels, have been gradually employed in medical image analysis. However, the quadratic computation complexity of attention mechanism has hindered the deployment of vision transformer in clinical practices. Bearing the analysis above in mind, this study introduces an integrated self-attention mechanism with both softmax and linear modules to guarantee efficiency and expressiveness, simultaneously. To be specific, a portion of query and key tokens, which are much less than the original query and key tokens, are adopted in the attention module by adding a set of proxy tokens. Note that the proxy tokens can fully utilize both the advantages of softmax and linear attention. To evaluate the performance of the presented approach, the comparison experiments between state-of-the-art algorithms and the proposed approach are conducted. Experimental results demonstrate that the proposed approach achieves superior outcome over the state-of-the-art algorithms on the publicly available datasets. Accordingly, the proposed approach can be taken as a potentially valuable instrument in clinical practices.

Label credibility correction based on cell morphological differences for cervical cells classification

Cervical cancer is one of the deadliest cancers that pose a significant threat to women’s health. Early detection and treatment are commonly used methods to prevent cervical cancer. The use of pathological image analysis techniques for the automatic interpretation of cervical cells in pathological slides is a prominent area of research in the field of digital medicine. According to The Bethesda System, cervical cytology necessitates further classification of precancerous lesions based on positive interpretations. However, clinical definitions among different categories of lesion are complex and often characterized by fuzzy boundaries. In addition, pathologists can deduce different criteria for judgment based on The Bethesda System, leading to potential confusion during data labeling. Noisy labels due to this reason are a great challenge for supervised learning. To address the problem caused by noisy labels, we propose a method based on label credibility correction for cervical cell images classification network. Firstly, a contrastive learning network is used to extract discriminative features from cell images to obtain more similar intra-class sample features. Subsequently, these features are fed into an unsupervised method for clustering, resulting in unsupervised class labels. Then unsupervised labels are corresponded to the true labels to separate confusable and typical samples. Through a similarity comparison between the cluster samples and the statistical feature centers of each class, the label credibility analysis is carried out to group labels. Finally, a cervical cell images multi-class network is trained using synergistic grouping method. In order to enhance the stability of the classification model, momentum is incorporated into the synergistic grouping loss. Experimental validation is conducted on a dataset comprising approximately 60,000 cells from multiple hospitals, showcasing the effectiveness of our proposed approach. The method achieves 2-class task accuracy of 0.9241 and 5-class task accuracy of 0.8598. Our proposed method achieves better performance than existing classification networks on cervical cancer.

Deep learning-based skin lesion analysis using hybrid ResUNet++ and modified AlexNet-Random Forest for enhanced segmentation and classification.

Skin cancer is considered globally as the most fatal disease. Most likely all the patients who received wrong diagnosis and low-quality treatment die early. Though if it is detected in the early stages the patient has fairly good chance and the aforementioned diseases can be cured. Consequently, diagnostic identification and management of the patient at this level becomes a rather enormous task. This paper offers a cutting-edge hybrid deep learning approach of better segmentation and classification of skin lesions. The proposed method incorporates three key stages: preprocessing, segmentation of lesions, and classification of lesions. By the stage of preprocessing, a morphology-based technique takes out hair so as to enhance the segmentation precision to use the cleansed images for subsequent analysis. Segmentation cuts off the lesion from the surrounding skin, giving the classification phase a dedicated area of interest and the ability to clear the background noise that may affect classification rates. The isolation enables the model to better analyze anatomical lesion features in order to achieve accurate benign and malignant classifications. Using ResUNet++, the cutting-edge deep learning architecture, we achieved accurate lesion segmentation. Next, we will modify and use an AlexNet-Random Forest (AlexNet-RF) based classifier for robust lesion classification. The proposed hybrid deep learning model is intensively validated on the Ham10000 data set which is one of the most popular datasets for skin lesions analysis. The obtained results show that the utilized approach, compared to the previous ones, is more effective, giving better segmentation and classification results. This method takes advantage of ResUNet++ strong classification skill and modified AlexNet-Random Forest robustness for more accurate segmentation. There is a high probability that ResUNet++, which is highly proficient at medical image segmentation, can produce better segmentation of lesions than the simpler models. The composition of AlexNet's extraction of features with Random Forest ability to reduce overfitting possibly may be more precise in the classification when compared to using only one model.

Dual Convolutional Neural Network for Skin Cancer Classification

Skin cancer detection through deep learning is an evolving field, where convolutional neural networks (CNNs) have proven to be very effective in feature extraction. However, this approach still faces some limitations due to the use of data augmentation, It is the generation of artificial images. Which significantly increase the computational load without generate new clinically meaningful data and may introduce shadowed features. Therefore, this study aims to propose a new approach that use CNNs to extract important features from skin cancer medical images using the HAM 10000 dataset. The proposed approach involves training two different CNN architectures, extracting features from convolutional layers, and then use PCA to make the retrieved features less dimensional. In order to categorize skin cancer into seven different categories of skin lesions, the remaining features are then merged and fed into a classifier that uses neural networks. In comparison to earlier studies that employed CNN architectures on the same dataset, the results demonstrated that this method preserves significant information while improving computational efficiency and achieving superior classification performance. The suggested approach achieved 95.66% accuracy for multi-class classification.

Application of Custom Ant Lion Optimization Convolutional Neural Networks for Liver Lesion Classification System

Application of Custom Ant Lion Optimization Convolutional Neural Networks for Liver Lesion Classification System

Leveraging Partial Labels for Cervical Lesion Classification via a Multilabel Approach

Leveraging Partial Labels for Cervical Lesion Classification via a Multilabel Approach

Prospective Evaluation of Accelerated Brain MRI Using Deep Learning-Based Reconstruction: Simultaneous Application to 2D Spin-Echo and 3D Gradient-Echo Sequences.

To prospectively evaluate the effect of accelerated deep learning-based reconstruction (Accel-DL) on improving brain magnetic resonance imaging (MRI) quality and reducing scan time compared to that in conventional MRI. This study included 150 participants (51 male; mean age 57.3 ± 16.2 years). Each group of 50 participants was scanned using one of three 3T scanners from three different vendors. Conventional and Accel-DL MRI images were obtained from each participant and compared using 2D T1- and T2-weighted and 3D gradient-echo sequences. Accel-DL acquisition was achieved using optimized scan parameters to reduce the scan time, with the acquired images reconstructed using U-Net-based software to transform low-quality, undersampled k-space data into high-quality images. The scan times of Accel-DL and conventional MRI methods were compared. Four neuroradiologists assessed the overall image quality, structural delineation, and artifacts using Likert scale (5- and 3-point scales). Inter-reader agreement was assessed using Fleiss' kappa coefficient. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated, and volumetric quantification of regional structures and white matter hyperintensities (WMHs) was performed. Accel-DL showed a mean scan time reduction of 39.4% (range, 24.2%-51.3%). Accel-DL improved overall image quality (3.78 ± 0.71 vs. 3.36 ± 0.61, P < 0.001), structure delineation (2.47 ± 0.61 vs. 2.35 ± 0.62, P < 0.001), and artifacts (3.73 ± 0.72 vs. 3.71 ± 0.69, P = 0.016). Inter-reader agreement was fair to substantial (κ = 0.34-0.50). SNR and CNR increased in Accel-DL (82.0 ± 23.1 vs. 31.4 ± 10.8, P = 0.02; 12.4 ± 4.1 vs. 4.4 ± 11.2, P = 0.02). Bland-Altman plots revealed no significant differences in the volumetric measurements of 98.2% of the relevant regions, except in the deep gray matter, including the thalamus. Five of the six lesion categories showed no significant differences in WMH segmentation, except for leukocortical lesions (r = 0.64 ± 0.29). Accel-DL substantially reduced the scan time and improved the quality of brain MRI in both spin-echo and gradient-echo sequences without compromising volumetry, including lesion quantification.

Enhanced Lesion Classification Based on YOLO Architectures Using Thermal Breast Images on a Patient by Patient Basis

Enhanced Lesion Classification Based on YOLO Architectures Using Thermal Breast Images on a Patient by Patient Basis

Erb palsy due to birth injury (a draft of the clinical recommendations)

The article presents a draft of clinical recommendations for the diagnosis and treatment of Erb palsy in birth injury. These recommendations are intended for physicians of maternity hospitals and are aimed at providing information support for decision-making by a doctor in the event of a birth injury, as well as improving the quality of medical care to a patient. The recommendations correspond to the latest scientific data on the topic, and also contain information that is of an applied nature for practical activities. These practical recommendations are offered for public discussion and are posted in full on the website of the Ministry of Health of the Russian Federation. The recommendations contain an information note on Erb palsy in birth injury, including the definition, etiology and pathogenesis of the disease. Three types of classification of generic traumatic lesions of the brachial plexus are presented: by prevalence, by severity of nerve damage and by severity of paresis. The clinical features of the disease are considered in detail. To diagnose the disease, an algorithm of physical examination is proposed, and the use of instrumental research methods: MRI, ultrasonic, X-ray and electromyographic studies. Recommendations for the conservative treatment of pathology and the use of physiotherapy and rehabilitation techniques are given. An algorithm of the doctor’s actions and an information certificate about the disease are proposed to inform the patient’s parents. The recommendations are given taking into account the level of credibility of the recommendations and the level of reliability of the evidence.

A Systematic Review of the Applications of Deep Learning for the Interpretation of Positron Emission Tomography Images of Patients with Lymphoma.

Background: Positron emission tomography (PET) is a valuable tool for the assessment of lymphoma, while artificial intelligence (AI) holds promise as a reliable resource for the analysis of medical images. In this context, we systematically reviewed the applications of deep learning (DL) for the interpretation of lymphoma PET images. Methods: We searched PubMed until 11 September 2024 for studies developing DL models for the evaluation of PET images of patients with lymphoma. The risk of bias and applicability concerns were assessed using the prediction model risk of bias assessment tool (PROBAST). The articles included were categorized and presented based on the task performed by the proposed models. Our study was registered with the international prospective register of systematic reviews, PROSPERO, as CRD42024600026. Results: From 71 papers initially retrieved, 21 studies with a total of 9402 participants were ultimately included in our review. The proposed models achieved a promising performance in diverse medical tasks, namely, the detection and histological classification of lesions, the differential diagnosis of lymphoma from other conditions, the quantification of metabolic tumor volume, and the prediction of treatment response and survival with areas under the curve, F1-scores, and R2 values of up to 0.963, 87.49%, and 0.94, respectively. Discussion: The primary limitations of several studies were the small number of participants and the absence of external validation. In conclusion, the interpretation of lymphoma PET images can reliably be aided by DL models, which are not designed to replace physicians but to assist them in managing large volumes of scans through rapid and accurate calculations, alleviate their workload, and provide them with decision support tools for precise care and improved outcomes.

Edge Artificial Intelligence Device in Real-Time Endoscopy for Classification of Gastric Neoplasms: Development and Validation Study.

We previously developed artificial intelligence (AI) diagnosis algorithms for predicting the six classes of stomach lesions. However, this required significant computational resources. The incorporation of AI into medical devices has evolved from centralized models to decentralized edge computing devices. In this study, a deep learning endoscopic image classification model was created to automatically categorize all phases of gastric carcinogenesis using an edge computing device. A total of 15,910 endoscopic images were collected retrospectively and randomly assigned to train, validation, and internal-test datasets in an 8:1:1 ratio. The major outcomes were as follows: 1. lesion classification accuracy in six categories: normal/atrophy/intestinal metaplasia/dysplasia/early/advanced gastric cancer; and 2. the prospective evaluation of classification accuracy in real-world procedures. The internal-test lesion-classification accuracy was 93.8% (95% confidence interval: 93.4-94.2%); precision was 88.6%, recall was 88.3%, and F1 score was 88.4%. For the prospective performance test, the established model attained an accuracy of 93.3% (91.5-95.1%). The established model's lesion classification inference speed was 2-3 ms on GPU and 5-6 ms on CPU. The expert endoscopists reported no delays in lesion classification or any interference from the deep learning model throughout their exams. We established a deep learning endoscopic image classification model to automatically classify all stages of gastric carcinogenesis using an edge computing device.

Beyond binary: multi-class skin lesion classification with AlexNet transfer learning-towards enhanced dermatological diagnosis

ProblemSkin lesions are the major indicator for diagnosing different skin diseases, which are caused by the abnormal growth of skin cells. Skin cancer, one of the most fatal types of cancer in the world, relies on the proper diagnosis of skin lesions and other relevant disease indicators for early detection, which can enhance treatment and increase life expectancy by up to 92%. Diagnosing skin lesions accurately is challenging, even for experienced practitioners.ChallengeArtificial Intelligence (AI) techniques based on deep learning have proved promising in the field of medical image processing and diagnosis. However, there is a lack of comprehensive medical image datasets. This paper will present a skin lesion classification model that employs a transfer learning approach based on Alex Net and classifies eight different skin diseases. This proposed model was trained, validated, and tested using ISIC 2019 challenge data with a very abnormal class distribution. This variation in class size has been overcome using data augmentation and resizing in the preprocessing phase. Preprocessing is performed in such a way that no colour detail or rich identification features will be lost from the images.ResultsThe proposed model achieved a testing accuracy of 90.9%, with a precision of 90.8%, a sensitivity of 90.9%, and an F1-Score of 90.8%. The values for precision and sensitivity are among the highest reported, specifically for models using the ISIC 2019 dataset.