L. Jani Anbarasi

Introduction: Leukemia is a type of cancer that originates in the bone marrow, causing uncontrolled production of abnormal white blood cells that disrupt normal blood function and weaken the immune system. Manual inspection is time-consuming and error-prone, relying heavily on the expertise and experience of medical professionals. Methods: The proposed study presents a hybrid model for classifying leukemia by integrating transfer learning and neutrosophic domain enhancement. Neutrosophic domain transformation splits the RGB channel image into Truth (T), Falsity (F), and Indeterminacy (I) components to address uncertainty, ambiguity, and poor contrast in blood cell representations. This enables the improvement of features more directly linked to leukemia identification. The images are augmented using wavelet sharpening and contrast-limited adaptive histogram equalization (CLAHE) on the T component, total variation minimization (TVM) on the F component, and wavelet shrinkage denoising on the I component. Results: This framework was trained and tested on the Leukemia Blood Cell Image Classification dataset, which included 3,256 peripheral blood smear (PBS) images across 4 classes: Benign, Early, Pre, and Pro. A transfer learning architecture based on MobileNetV2 was used for classification, and training was conducted using a 70:15:15 split for training, validation, and testing, respectively. The proposed neutrosophic-enhanced MobileNetV2 model achieved an overall testing accuracy of 98.36% and a macro F1-score of 0.98, demonstrating significant enhancement in multi-class leukemia classification. Discussions: The incorporation of the neutrosophic enhancement method significantly improves classifier performance, resulting in higher accuracy without increasing computational power.

Introduction: Accurate segmentation of kidney masses and structure is essential for medical application including diagnosis and treatment. This research proposed the dual track hybrid VHUCS-Net architecture which effectively highlights structural size-shape variants, boundaries and complex structural features in kidney disease. Methods: Efficient segmentation is achieved by integrating the transformer enhanced U-Net model with the contrast optimized Protuberance Detection Network (PDN) model. The process begins with analysing kidney images using a standard U-Net combined with Vision Transformer attention and a High Resolution Network (HRNet) which capture global dependencies while preserving high resolution features resulting in accurate segmentation of the kidney region. Also, the masked kidney image undergoes processing through a contrast optimized PDN model with multi scale pooling, contrast enhancement, boundary refinement and explicit feature fusion to segment the mass region thereby enhancing mass localization improving border identification and enabling accurate abnormality detection. The resulting features are fused to provide a refined mass segmentation result that exactly identifies the location and structural abnormalities. Results: The VHUCS-Net model was evaluated using the kidney segmentation dataset achieving an intersection over union score of 0.9441 and a dice coefficient of 0.9712 showing outstanding segmentation precision. Discussion: These results indicate improved diagnostic efficiency and support clinical decision making by providing more accurate and interpretable segmentation outputs. Moreover, VHUCS-Net is validated with additional publicly available datasets with image mask correspondence, therefore proving the model effectiveness and generalizability across many segmentation tasks. The results highlight the capability of the proposed VHUCS-Net model to enhance diagnostic accuracy and assist clinical decision making through more detailed and interpretable segmentation outcomes.

Introduction: Breast cancer, one of the most life-threatening diseases that commonly affects women, can be effectively diagnosed using breast ultrasound imaging. A hybrid deep learning based ensemble framework combining the effectiveness of different convolutional neural network models has been proposed for breast ultrasound image classification. Methods: Three distinct deep learning models, namely, EffcientNetB7, DenseNet121, and ConvNeXtTiny, have been independently trained on breast ultrasound image datasets in parallel to capture complementary representations. Local features are extracted using EffcientNetB7 through depthwise separable convolutions, whereas structural details are preserved by DenseNet121 utilizing dense connectivity. Global spatial relationships are modeled using ConvNeXtTiny via large kernel operations. Diverse local, global, and hierarchical features extracted with respect to multiple perspectives are integrated into a high-dimensional unified representation from which non-linear decision boundaries are learned utilizing XGBoost as the feature fusion classifier. Additionally, a soft voting ensemble method averages the predicted probabilities of the individual convolutional network architectures. Results: The model was evaluated using the BUSI dataset, the BUS-UCLM dataset, and the UDIAT dataset. The accuracy, precision, recall, F1 score, and AUC values obtained on the BUSI data set are 88.46%, 88.49%, 88.46%, 88.45%, and 95.38%, respectively. On the BUS-UCLM dataset, the corresponding values are 90. 51%, 90. 56%, 90. 51%, 90. 51%, and 96. 23%, respectively. The accuracy, precision, recall, F1 score, and AUC values obtained on the UDIAT dataset are 96.97%, 100.00%, 90.91%, 95.24%, and 99.17%, respectively. The decision-making capability of the model has been highlighted using SHAP and Grad-CAM visualizations, further improving the interpretability and transparency of the model, and making it more robust for breast ultrasound image classification. Discussion: The HED-Net framework exhibits significant potential for clinical application by enhancing diagnostic accuracy and decreasing interpretation time, particularly in resource-limited environments where expert radiologists are in short supply.

The identification of cephalometric landmarks is an essential and imperative process commonly employed in the planning and diagnosis of orthodontic treatments. Computer-aided completely automated methods have the potential to enhance the efficiency of orthodontists and orthognathic surgeons by accurately identifying landmarks from cephalograms. In this paper, an AdamW Golden Search Optimization based multi-level attention enhanced Stacked Feature Generator architecture (AWGSO_SFG_CephX) is proposed to automatically predict 19 landmarks from cephalometric radiographs. The Stacked Feature Generator (SFG_CephX) model utilized multiple multi-level attention integrated feature extractors combined with a characteristic meta-learner to provide well-balanced landmark prediction results with higher Successful Detection Rates (SDR_s) and minimum Mean Radial Errors (MRE_s) with Standard Deviations (SD_s). The proposed architecture collaborated gaussian dropout layers to acquire more resilient unique landmark features that are not excessively reliant on specific input landmarks and spatial dropout strategy for learning more robust landmark features from the skull X-rays by randomly dropping out the feature maps during the training. The SFG_CephX is trained using the AWGSO technique by integrating Golden Search Optimization (GSO) algorithm with AdamW optimizer. The proposed framework achieved average SDRs of 89.17%, 79.8% for Test1 and Test2 dataset in 2mm precision using the Institute of Electrical and Electronics Engineers (IEEE) organized 2015 International Symposium on Biomedical Imaging (ISBI) grand challenge for dental X-ray analysis dataset. The proposed AWGSO_SFG_CephX is evaluated using a clinical private dataset consisting of 130 skull X-rays obtained from Solanki Dental Care Clinic in Sharjah, United Arab Emirates, and obtained an average SDR of 75.71% within precision range of 2mm.

Introduction: Lung cancer remains as an important source of cancer-related mortality worldwide, demonstrating a substantial challenge to public health systems. The absence of evident symptoms in the early stages makes timely diagnosis of lung cancer challenging. Early identification and treatment will reduce the mortality rate caused by lung cancer. Abnormal growths identified as lung or pulmonary nodules can be found in the lungs and some of these could be malignant. A Computer-Aided Detection (CAD) framework can aid in identifying pulmonary nodules by investigating medical images. Automated CAD systems assist radiologists by reducing the diagnostic workload and increasing the possibility of early lung cancer identification. Finding and accurately outlining lung nodules is the specific task of lung nodule segmentation in medical image analysis. Methods: Multi-scale UNet, Feature Pyramid Network (FPN) with Linear Attention Mechanism and UNet with Asynchronous Convolution Blocks (ACB) and Channel Attention Mechanism were used to segment lung nodules. Multi-scale UNet improvises the traditional UNet architecture by incorporating multi-scale convolutional operations, which improves feature extraction and boosts segmentation accuracy. The UNet with ACB and Channel Attention Mechanism employs a cross-like receptive field that can reduce the impact of redundant information in obtaining representative characteristics. FPN with Linear Attention mechanism uses a multi-scale feature pyramid to identify nodules of different sizes and a linear attention mechanism is employed to improve feature extraction. FPN with Linear Attention mechanism attains a linear time and spatial complexity while effectively segmenting pulmonary nodules. Results and discussion: Employing the FPN with Linear Attention mechanism yielded the highest performance in the experiments. The highest results in the study using FPN with Linear Attention were achieved using GELU on the LIDC-IDRI dataset with a DSC of 71.59% and IoU of 58.57%. The smooth, probabilistic weighting of GeLU complements the model's attention mechanisms.

Introduction: Breast cancer is a severe illness predominantly affecting women, and in most cases, it leads to loss of life if left undetected. Early detection can significantly reduce the mortality rate associated with breast cancer. Ultrasound imaging has been widely used for effectively detecting the disease, and segmenting breast ultrasound images aid in the identification and localization of tumors, thereby enhancing disease detection accuracy. Numerous computer-aided methods have been proposed for the segmentation of breast ultrasound images.Methods: A deep learning-based architecture utilizing a ConvMixer-based encoder and ConvNeXT-based decoder coupled with convolution-enhanced multihead attention has been proposed for segmenting breast ultrasound images. The enhanced ConvMixer modules utilize spatial filtering and channel-wise integration to efficiently capture local and global contextual features, enhancing feature relevance and thus increasing segmentation accuracy through dynamic channel recalibration and residual connections. The bottleneck with the attention mechanism enhances segmentation by utilizing multihead attention to capture long-range dependencies, thus enabling the model to focus on relevant features across distinct regions. The enhanced ConvNeXT modules with squeeze and excitation utilize depthwise convolution for efficient spatial filtering, layer normalization for stabilizing training, and residual connections to ensure the preservation of relevant features for accurate segmentation. A combined loss function, integrating binary cross entropy and dice loss, is used to train the model.Results: The proposed model has an exceptional performance in segmenting intricate structures, as confirmed by comprehensive experiments conducted on two datasets, namely the breast ultrasound image dataset (BUSI) dataset and the BrEaST dataset of breast ultrasound images. The model achieved a Jaccard index of 98.04% and 94.84% and a Dice similarity coefficient of 99.01% and 97.35% on the BUSI and BrEaST datasets, respectively.Discussion: The ConvMixer and ConvNeXT modules are integrated with convolution-enhanced multihead attention, which enhances the model's ability to capture local and global contextual information. The strong performance of the model on the BUSI and BrEaST datasets demonstrates the robustness and generalization capability of the model.