Sarvajanik College of Engineering and Technology

An affordable and sustainable way to meet the growing demand for fresh water worldwide is to purify saline water using square pyramid solar still (SqPySS). The primary drawback to the SqPySS's global application is its lower distillate output. This research study investigates the enhancement of distillate output in SqPySS by combining hollow fins embedded with phase changing material (PCM) in organic nanofluid blended saline water. In the first stage of experiment, the effect of square fin (SqFn) and circular fin (CirFn) filled with paraffin wax were evaluated for 1 and 2 cm depth of pure saline water. Based on the outcome of the first experiments, the second stage of experiment was performed at constant depth of Chitosan organic nanofluid blended saline water using both PCM embedded fins. Both stages of experiments were carried out at Surat, Gujarat, India (21.1702 degrees N, 72.8311 degrees E) for three consecutive days. The maximum distillate of 3430 ml/m2 center dot day was obtained for CirFn with the 1 cm water depth in the first stage with an efficiency of 32.02%. While in the second stage of experiments, the highest yield of 4056 ml/m2 center dot day with an efficiency of 35.31% was found for CirFn with the 1 cm depth of Chitosan blended saline water. The distillate yield and efficiency enhancement of 18.25% and 10.27% for CirFn, while 30.58% and 10.93% for SqFn was obtained in second stage of experiments. The higher thermal conductivity, extended heat transfer area and prolonged heat storage plays a significant role in performance enhancement of augmented SqPySS.

White blood cell (WBC) segmentation of microscopic blood smear images (MBSI) is a significant process in the diagnosis of blood-related diseases like anemia, leukemia, and parasites. Traditional manual segmentation is tiresome and sensitive to inter-observer variation, which limits its application to a big-scale clinical process. To address these issues, this paper presents a hybrid DINO-Net approach, which works on WBC segmentation. The suggested solution uses a DINO-Net encoder with self-supervised training to obtain discriminative features and context-sensitive features, which are then decoded by a U-Net-based network to obtain the correct segmentation masks. The effectiveness of the offered framework is contrasted with popular segmentation models, i.e., U-Net, DeepLabv3+, and a DINOv2-based one. The experimental outcomes demonstrate that the suggested strategy is always more successful in comparison to the comparison models during the training, validation, and testing stages. Precisely, Dice coefficients of $93.00 \%, 89.25 \%$, and 90.35 %, as well as IoU scores of $87.10 \%, 81.90 \%$, and 85.23 %, are obtained, respectively. Also, the recall values of $92.24 \%, 80.35 \%$, and 81.64 % show that WBC regions are reliably detected. These results prove that self-supervised transformer-Convolutional Neural Network (CNN) addition makes it a possibility to segment WBC.

Wastewater from kraft paper mills contains high concentrations of volatile fatty acids (VFAs), which contribute significantly to odour pollution and environmental degradation. Effective VFA removal is crucial for improving the quality of wastewater and minimizing its ecological impact. This study compares the performance of three aeration methods—mechanical aeration (MA), diffused air aeration (DAA), and combined diffused air and mechanical aeration (CDAMA)—for reducing VFAs, odour intensity, and key wastewater parameters such as chemical oxygen demand (COD), biological oxygen demand (BOD), and pH. A laboratory-scale aeration system (6-L capacity) was used to evaluate the treatment efficiency of each method. Among the methods tested, CDAMA achieved the highest removal efficiency, reducing VFAs, odour intensity, COD, and BOD by 65.9

Heart disease remains one of the leading causes of cardiovascular mortality worldwide, creating a persistent challenge for healthcare systems. Early diagnosis and intervention are critical for minimizing complications, improving patient outcomes, and reducing mortality rates. Electrocardiography (ECG) is a widely used, non-invasive diagnostic technique that provides valuable insights into cardiac rhythm, myocardial function, and coronary artery health. However, manual interpretation of ECG images is often hindered by complexity, inter-patient variability, and the large volume of data generated. Recent advances in Artificial Intelligence (AI), particularly deep learning techniques such as Convolutional Neural Networks (CNNs), have shown significant promise in automating heart disease detection. This study presents an AI-powered deep learning model that employs CNNs to classify and predict the likelihood of heart disease from ECG images. To improve accuracy and computational efficiency, the research integrates SqueezeNet as a pre-trained transferable model. The proposed methodology includes dataset selection, pre-processing, and augmentation, followed by training, validation, and performance evaluation. Key metrics such as accuracy, precision, recall, F1-score, and confusion matrix are used for assessment, with visualizations provided to demonstrate model effectiveness and generalization capability. Experimental results confirm the potential of this approach in achieving high predictive accuracy, reducing reliance on manual diagnosis, and supporting clinicians in early disease detection. This work highlights the effectiveness of deep learning in ECG-based heart disease prediction and its potential integration into clinical decision support systems. The model offers a scalable and reliable tool for enhancing diagnostic efficiency, enabling timely interventions, and ultimately contributing to reduced cardiovascular mortality.

Accurate diagnosis and grading of cancer involve detecting and classifying mitotic cells within histopathological Whole Slide Images (WSIs), which is a critical but challenging task for pathologists. In the past decade, many deep learning-based approaches have been introduced; however, these approaches have limitations such as poor mitotic region detection, misclassification due to structural similarity, requirement of highly annotated data, and difficulty in working with different stain protocols. Hence, we propose a novel transformer-driven framework that incorporates a Real-Time Detection Transformer (RT-DETR) for the detection of mitotic cells and a Vision Transformer (ViT) for the classification of mitotic cells. The RT-DETR model supports accurate and efficient localization of mitotic cells as a result of its efficient detection capabilities. The ViT model supports fine-grained classification of mitotic cells by modeling global context among visual features. The proposed framework was evaluated against three WSI datasets. The proposed framework outperforms the existing YOLOv8-based Mitosis Detection and Classification (MDC) approach. The results of the experiments show that the proposed framework achieves 40.56% higher accuracy, 16.15% higher precision, 21% higher recall and 47% higher F1-score as compared to HR-YOLOv8 [11]. The framework also provides better results against other experiments carried out using three state-of-the-art models, viz, YOLOv12, RT-DETR, and YOLOv12 integrated with ViT. The proposed transformer-driven framework provides robust detection and accurate classification of mitotic cells in WSI.