Metaheuristic optimization algorithms have gained wide adoption in engineering and scientific domains. However, many swarm-based methods struggle to balance exploration and exploitation, often converging prematurely on suboptimal solutions. The Sand Cat Swarm Optimization (SCSO) algorithm is one such method, with limited exploration ability constraining its performance on complex problem landscapes. This paper introduced the Enhanced Sand Cat with Selective Opposition (ESCSO) algorithm which combines opposition-based learning with a velocity mechanism to overcome this limitation. In ESCSO, under-performing candidates referred to as Sigma-variant cats are identified using Spearman correlation and replaced with their opposite solutions to inject diversity into the search process. Stronger candidates termed Sigma cats, act as elite guides pulling the search toward better regions. A PSO-inspired velocity update governs both roles, keeping exploration and exploitation in balance rather than letting one dominate. Tested across 30 benchmark functions plus two real engineering problems, reflectarray antenna design and microgrid energy management, ESCSO achieves competitive convergence, solution quality, and robustness when compared to recent state-of-the-art methods.

Indoor object detection presents unique challenges such as occlusions, varying lighting conditions, and cluttered environments. While several object detection frameworks, including RetinaNet, Faster R-CNN, SSD, and EfficientDet, have been proposed, they often suffer from high computational cost, reduced inference speed, and limited accuracy in terms of mean Average Precision (mAP), particularly in real-time scenarios. In this study, lightweight YOLO variants, namely YOLOv7, YOLOv8s, YOLOv9s, and a fine-tuned YOLOv9s which considers the optimized training strategy based on albumentations. All the models are evaluated for indoor object detection using the RGB TUT Indoor dataset. The models are assessed using precision, recall, mAP@0.5, and mAP@0.5:0.95. The experimental results demonstrate that the fine-tuned YOLOv9s consistently outperforms the baseline YOLOv9s model across all evaluation metrics, confirming the effectiveness of proposed training optimizations. Specifically, the fine-tuned YOLOv9s achieves a precision of 97.9%, recall of 96.1%, mAP@0.5 of 99.1%, and mAP@0.5:0.95 of 88.7%. These improvements highlight the impact of systematic training refinement beyond standard model configuration. Among the evaluated models, YOLOv8s achieves the highest inference speed of 90 FPS in 11.1 ms, making it suitable for ultra-low-latency applications such as smart homes, assistive systems, and robotics. In contrast, the fine-tuned YOLOv9s provides a superior balance between accuracy and efficiency, making it more suitable for accuracy-sensitive indoor environments where detection reliability is critical. Overall, the study demonstrates that carefully optimized training strategies can significantly enhance the performance of YOLOv9s without architectural modifications, providing practical insights for real-time indoor object detection systems.

The rising complexity of cyber threats has underscored the critical need for intelligent Intrusion Detection Systems (IDS) that can efficiently handle high-dimensional data and evolving attack patterns. Traditional IDS often suffer from redundancy and inefficiencies. This research integrating Support Vector Machine (SVM) and C4.5 decision tree models with metaheuristic optimized machine learning framework for intrusion detection, utilizing the NSL-KDD and UNSW-NB15 datasets. The proposed methodology involves comprehensive data preprocessing, discriminative feature extraction, and hybrid model development using various metaheuristic algorithms, including particle swarm optimization (PSO), modified PSO, BAT, Cuckoo Search (CS), Ant Colony Optimization (ACO), Grey Wolf Optimizer (GWO), Harris Hawks Optimization (HHO), and Genetic Algorithm (GA). These algorithms optimize the selection of relevant features to enhance detection accuracy and reduce false positives. Experimental analysis reveals that hybrid models particularly MPSO-C45, CS-C45, BAT-SVM, and MPSO-SVM significantly outperform traditional classifiers in both datasets. On NSL-KDD, MPSO-C45 and CS-C45 achieved near-perfect metrics F-measure of 0.993, sensitivity of 0.994 with low runtime of 1.7 s. While on UNSW-NB15, GWO-C45, CS-C45, and PSO-SVM achieved high classification results accuracy of 1.000, and MSE of 0. Comparative evaluation against existing state-of-the-art approaches confirms the superiority of the proposed models, particularly in terms of balancing accuracy, runtime, and false positive reduction. However, GA-based models consistently underperformed, indicating limitations in their optimization capability. Despite excellent results, the study acknowledges challenges in real-time deployment and computational overhead.

In recent years, metaheuristic algorithms have gained prominence as powerful computational techniques for solving complex optimization problems across various domains, including healthcare. Their ability to effectively explore vast solution spaces enables them to identify optimal or near-optimal solutions to challenging problems. Despite their growing importance, metaheuristic algorithms, inspired by natural and human problem-solving strategies, are increasingly applied in healthcare to address complex optimization challenges such as diagnosis, treatment planning, and resource allocation. But traditional ML models often face problems like getting stuck in training, taking too much time to compute, and not balancing exploration and exploitation properly. These issues make them less effective for solving complex medical classification problems. To overcome these limitations, this study introduces new hybrid metaheuristic models such as Ropalidia Marginata (RM) hybrid with various metaheuristic algorithms such as Ant Colony Optimization (ACO), particle swarm optimization (PSO), Firefly, Artificial Bee Colony (ABC), Grey Wolf Optimizer (GWO), Differential Evolution (DE), and Bat Algorithm. Further all these hybrid metaheuristic algorithms are combined with a Feedforward Neural Network (FFNN) to train the network more efficiently. The proposed method uses the dominance-based behavior of RM wasp to improve the global search ability, is stable during learning, and improves classification accuracy. The performance of proposed frameworks is to check against several start of the art models on three critical medical datasets such as breast cancer, diabetes, and heart disease. The results showed that proposed models gave better results than the traditional algorithms in terms of accuracy, mean squared error (MSE), standard deviation (SD), and AUC-ROC.

Indoor object detection presents unique challenges such as occlusions, varying lighting conditions, and cluttered environments. While several object detection frameworks, including RetinaNet, Faster R-CNN, SSD, and EfficientDet, have been proposed, they often suffer from high computational cost, reduced inference speed, and limited accuracy in terms of mean Average Precision (mAP), particularly in real-time scenarios. In this study, lightweight YOLO variants, namely YOLOv7, YOLOv8s, YOLOv9s, and a fine-tuned YOLOv9s which considers the optimized training strategy based on albumentations. All the models are evaluated for indoor object detection using the RGB TUT Indoor dataset. The models are assessed using precision, recall, mAP@0.5, and mAP@0.5:0.95. The experimental results demonstrate that the fine-tuned YOLOv9s consistently outperforms the baseline YOLOv9s model across all evaluation metrics, confirming the effectiveness of proposed training optimizations. Specifically, the fine-tuned YOLOv9s achieves a precision of 97.9%, recall of 96.1%, mAP@0.5 of 99.1%, and mAP@0.5:0.95 of 88.7%. These improvements highlight the impact of systematic training refinement beyond standard model configuration. Among the evaluated models, YOLOv8s achieves the highest inference speed of 90 FPS in 11.1 ms, making it suitable for ultra-low-latency applications such as smart homes, assistive systems, and robotics. In contrast, the fine-tuned YOLOv9s provides a superior balance between accuracy and efficiency, making it more suitable for accuracy-sensitive indoor environments where detection reliability is critical. Overall, the study demonstrates that carefully optimized training strategies can significantly enhance the performance of YOLOv9s without architectural modifications, providing practical insights for real-time indoor object detection systems.

Accurate and efficient disease diagnosis remains a critical challenge in the healthcare sector. With the growing availability of biomedical data, machine learning techniques have become invaluable tools for developing intelligent disease detection systems. Researchers have applied various algorithms, including artificial neural networks (ANNs), to improve classification accuracy. To further improve ANN performance, various optimization methods are applied to enhance learning and avoid the local minima problem, as each model demonstrates distinct performance characteristics. Therefore, this paper presents a hybrid Bio inspired Ropalidia Marginata Optimization-based hybrid neural network (RMO-NN) aimed at improving medical data classification. The proposed RMO-NN incorporates biologically inspired task allocation and dominance hierarchy mechanisms from RMO to optimize neural network learning performance effectively and reducing classification errors. To validate its effectiveness, the RMO-NN is tested on three large-scale medical datasets such as breast cancer, diabetes, and blood transfusion datasets and three medical images datasets. The performance of the proposed model is compared against two established metaheuristic neural models: Cuckoo Search Neural Network (CSNN) and Artificial Bee Colony Neural Network (ABCNN). The proposed RMO-NN model outperforms CSNN and ABCNN in terms of accuracy, MSE, SD, and convergence speed. And for medical images datasets the proposed is further validated with various start of art deep learning models. The results highlight the proposed model perform better on biomedical data classification tasks. The Proposed method significantly outperforms baseline approaches, achieving substantial accuracy, while introducing a novel RMO algorithm.

Car classification is important in daily life because there are many distinct types of automobiles made by various manufacturers. Although there are numerous methods for classifying autos, machine learning technologies have not been widely utilized, resulting in low accuracy levels. The goal of this paper is to create a machine learning system that is especially made to categories models of two Pakistan's top automakers, Toyota, and Honda. Ten Toyota models such as Avalon, Land Cruiser, Camry, Corolla, C-HR, Highlander, Prius, Tundra, RAV4, and Yaris and a dataset of Honda automobiles, which also includes 10 models (Accord, Civic, CR-V, Fit, HR-V, Insight, Odyssey, Passport, Pilot, and Ridgeline), are used to evaluate the model's performance. A deep learning-based VGG integrated with support vector machine (SVM) is proposed, utilizing a dataset from Kaggle.com, providing high-definition images for multiple classes. Comparisons with other models such as VGG16, AlexNet, and Convolutional Neural Network (CNN) reveal that the suggested model (VGG16 + SVM) achieves superior accuracy. For the Toyota dataset, the proposed model achieves 99% accuracy, outperforming VGG16 (66%), AlexNet (52%), and CNN (65%). Similarly, for the Honda dataset, the suggested model achieves 98% accuracy, surpassing VGG16 (96%), AlexNet (71%), and CNN (82%). In conclusion, the proposed deep learning-based model demonstrates enhanced accuracy in classifying Toyota and Honda cars, highlighting its effectiveness for image-based classification tasks in the automotive domain.

This research introduces an AI-driven outdoor object detection system aimed at enhancing navigation for visually impaired individuals (VIIs). VIIs often face significant challenges in accessing and interpreting visual information. Recent advancements in computer hardware and deep learning techniques have led to notable progress in developing assistive technologies for VIIs. However, existing datasets often focus on single scenarios and lack sufficient annotations to represent the diverse obstacles encountered in real-world settings. This limitation hinders the development of comprehensive object detection systems tailored to the needs of VIIs. The system utilizes advanced models such as YOLOv8 (Nano, Small, Medium), YOLOv9c, and YOLOv10n, with neighbor coordinates and C2FCIB attention modules trained on the WOTR dataset, which includes 20 classes of common outdoor objects. A comparative study evaluated the performance of these models across key metrics. The YOLOv8m model demonstrated balanced performance with an accuracy of 85.53. YOLOv8n showed slightly lower performance, with an accuracy of 77.05%, and the YOLOv8s model recorded an accuracy of 84.99%, precision and recall of 0.99, 0.89, matching YOLOv8n with an F1 score of 0.74. Similarly, YOLOv9c achieved an accuracy of 79.82%, and the proposed YOLOv10n model with neighbor coordinates and C2FCIB attention modules led with the highest accuracy of 89.33%, precision and recall of 0.99, 0.92, with F1 score of 0.79. A comparative analysis revealed that the proposed YOLOv10n with neighbor coordinates and C2FCIB attention modules achieved the highest accuracy with precision and recall indicating its reliability for assistive applications. In the realm of assistive technologies, similar AI-powered devices have been developed to aid visually impaired individuals. These innovations, alongside the described object detection system, exemplify the potential of AI in creating inclusive solutions that empower visually impaired individuals to navigate their environments more safely and independently.

Researchers have long been captivated by the intricate workings of the human brain, an enduring enigma. Extensive efforts have been devoted to unraveling its complexities, with disciplines like psychology employing experimentation and analysis to scrutinize and formulate models of brain function. Comprising billions of interconnected neurons, the human brain has inspired experts in deep learning to construct artificial neural networks capable of tasks akin to human brain functions, such as pattern and speech recognition. Despite substantial progress in artificial intelligence, advancements in memory storage capabilities have been relatively constrained. This study aims to investigate mechanisms for simulating long-term declarative episodic memory, reminiscent of human cognition, using one-shot deep-learning neural networks. The proposed deep learning architecture extends to Rosenblatt's C-system memory model, and experiments were conducted to assess the effectiveness of various adaptations of the C-system storage mechanism. The fashion MNIST dataset is used in the experiments, and the results indicate that these models exhibit proficient recall abilities, even when faced with a large number of input images. Furthermore, the study delves into emulating the forgetting process of the human brain. The experiment demonstrates that as the units in the C-system increase, the corresponding results also increase. Specifically, when employing 40,000 units, the system maintains an accuracy exceeding 92% for the sequence of images.

The current methods, while in use, continue to grapple with accuracy and effectiveness concerns. It is imperative to establish dependable solutions capable of distinguishing and categorizing people according to their assistive devices to tackle these issues. People with disabilities, such as those experiencing paralysis, limb deficiencies, or amputations, may encounter issues related to discrimination and inadequate support. Hence, this research was undertaken to detect and track people with conditions like paralysis, limb deficiency (Amelia), or amputation among the differently-abled population. Earlier investigations have predominantly focused on recognizing people and their mobility aids, utilizing a variety of methods such as Fast R-CNN, Faster R-CNN, RGB or RGB-D cameras, Kalman filters, and hidden Markov models. Modern deep learning models, including YOLO (You Only Look Once) and its variations, have gained substantial acceptance in current applications owing to their distinctive architectural designs and performance attributes. In this study, a substantial dataset comprising 4,300 images and 8,447 labels spanning five distinct categories is employed to assess the efficacy of YOLOv8, YOLOv5, and YOLOv7 models in the identification of people with disabilities. The evaluation findings show that YOLOv8, which achieved an overall precision of 0.907, performs better than both YOLOv5 (precision: 0.885) and YOLOv7 (precision: 0.906). Notably, YOLOv8 has the best wheelchair detection precision (0.998). Furthermore, YOLOv8 outperforms YOLOv5 (recall: 0.887) and YOLOv7 (recall: 0.925) in terms of recall performance (recall: 0.943). YOLOv8 achieves the greatest mean average accuracy (mAP@.5) value of 0.951, followed by YOLOv5 (mAP@.5: 0.942), and YOLOv7 (mAP@.5: 0.954). In a similar vein, of the three models, YOLOv8 has the best performance (mAP@.5:.95: 0.713). The analysis of detection time also shows that YOLOv8 performs best, processing 5,597 frames in just 5.9 milliseconds and achieving a remarkable frame rate of 169.49 frames per second.

Case Presentation:We present the case of a six-year-old child with autism who presented with persistent vomiting in the setting of a closed head injury (CHI). Computed tomography of the head was normal, but due to persistent vomiting a radiograph of the abdomen was done, which showed multiple, rare-earth magnets in the abdomen. There was no history of witnessed ingestion. These magnets had caused enteroenteric fistula formation leading to persistent vomiting.Discussion:In the setting of CHI, vomiting can be a sign of concussion or intracranial hemorrhage. In cases of CHI where intracranial pathology is ruled out and vomiting still persists, it is important to explore intra-abdominal causes of vomiting, especially in developmentally challenged children as they have higher incidence of unwitnessed foreign body ingestions.

Recent advancements in image processing and computer vision have brought significant transformations in healthcare technology, leading to significant improvements in diagnosis accuracy, cost-effectiveness, and time efficiency. Magnetic Resonance Imaging (MRI) is employed by the radiologist for its remarkable ability to detect even the most subtle brain abnormalities. This study considers a comprehensive analysis of the two prominent object identification frameworks, YOLOv5 and YOLOv7, leveraging state-of-the-art deep learning architectures to classify and detect brain cancers within MRI. The brain tumor dataset encompasses three distinct classes, including meningiomas, gliomas and pituitary tumors. To ensure precise segmentation of the tumor regions, the preprocessing phase incorporates advanced mask alignment techniques. This preprocessed dataset has been used to evaluate the performance of the deep learning models for brain tumor detection and classification. From the numerical results of YOLOv5, it was noticed that a recall score of 0.905 for box detection and 0.906 for mask segmentation, with a precision score of 0.94 and 0.936, respectively. At an IoU threshold of 0.5, both box detection and mask segmentation achieve a mAP of 0.947, whereas, at an IoU threshold of 0.5 to 0.95, they achieve mAPs of 0.666 and 0.657, respectively. In comparison, YOLOv7 exhibits strong performance with box detection accuracy of 0.936 and a mask segmentation accuracy of 0.935. The recall score are 0.904 for box detection and mask segmentation is 0.903. Notably, the mAP result at the IoU threshold of 0.5 are 0.94 for box detection and mask segmentation is 0.941. Over the broader IoU spectrum of 0.5 to 0.95, the mAP was 0.677 for box detection and 0.659 for mask segmentation. To underscore the novelty of the approach, the performance of the proposed framework is systematically compared with established methods such as RCNN, Faster RCNN, and Mask RCNN.

These days, mobile phones are the most commonly purchased goods. Thousands of new models with improved features, designs, and specifications are released yearly. An autonomous mobile price prediction system is required to assist customers in determining whether or not they can afford these devices. Many machine learning models exhibit varying performance degrees based on their architecture and learning properties. Ten widely used classifiers were assessed in this study: Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Decision Tree (DT), Naïve Bayes (NB), Linear Discriminant Analysis (LDA), AdaBoost, and Light Gradient Boosting (LGB). The F1-score, recall, accuracy, and precision of these models were evaluated. According to the findings, the results indicated that LR, with its use of the Elastic Net parameter, outperformed the others with 96% accuracy, 97% precision, 94% recall, and 96% F1-score. Other models like XGBoost, LGB, and SVM also showed strong performance, whereas KNN had the poorest performance. The study highlights the importance of selecting the appropriate model for accurate mobile price prediction. Among all the machine learning used in this paper, the LR classifier outperforms the other state-of-the-art models because of the elastic Net parameter used for mobile phone price prediction.

The Radial Basis Function Neural Network (RBFNN) is frequently employed in artificial neural networks for diverse classification tasks, yet it encounters certain limitations, including issues related to network latency and local minima. To tackle these challenges, researchers have explored various algorithms to enhance learning performance and alleviate local minima problems. This study introduces a novel approach that integrates the Crow Search Algorithm (CSA) with RBFNN to augment the learning process and address the local minima issue associated with RBFNN. The study evaluates the performance of this innovative model by comparing it to state-of-the-art models like Flower-pollination-RBNN (FP-NN), Artificial Neural Network (ANN), and the conventional RBFNN. To assess the efficacy of the proposed model, the study employs specific datasets, such as the Breast Cancer and Thyroid Disease datasets from the UCI Machine Repository. The simulation results illustrate that the proposed model surpasses other models in terms of accuracy, exhibiting lower Mean Squared Error (MSE) and Mean Absolute Error (MAE) values. Specifically, for the Breast Cancer dataset, the proposed model attains an accuracy of 99.9693%, MSE of 0.000307024, and MAE of 0.00789449. Likewise, for the Thyroid Disease dataset, the proposed model achieves an accuracy of 99.9535%, along with MSE of 0.000464932 and MAE of 0.0057098. For the diabetes dataset, the proposed model demonstrates an accuracy of 98.8073%, MSE of 0.003024, and MAE of 0.009449. In summary, this analysis underscores the enhanced accuracy and effectiveness of the proposed model when compared to traditional approaches.

The advent of information technology re-innovates all sectors of bio-sciences. Researchers use Semantic Web to improve web searching, mining and integration, which alleviates the time-consuming task of finding relevant and high-quality content. Semantics is improved through ontology engineering in any domain. Amended and developed ontologies will be uploaded to existing standardised and approved biomedical repositories. The establishment of a World Wide Web Consortium (W3C) approved and standardised ontology repository is the most ambitious goal. This work will solely focus on some selected agri-ontologies. The main objective is to promote outcome-based research and transformation styles of relevant expertise sharing. The intended goal is to win project funding to train and equip students with relevant skills and expertise. Need-based and market-oriented training and professional grooming are a tangible asset for students. The majority of traditional Web development freelancers are unaware of ontology or semantic web market demand. Freelancing is another option for expert Ontology developers. However, agriculture students are used to all the research vocabulary and terminologies in their area, but they do not know how to contribute their expertise to improve the efficiency of the Semantic Web in their domain. If the improvement in relevant ontology becomes a part of the Semantic Web, then it is termed 'Real-time Web semantics enhancement'. In other words, the target ontology becomes a part of the future Web of meaning.

In today's world, a large section of the human population suffers from treatable diseases such as heart problems, diabetes, skin cancer, stroke, liver disease, Parkinson's, malaria, and brain tumors. But due to lack of accessible and affordable healthcare the conditions may not be accurately or timely diagnosed that may lead to severe consequences, including disability or even death. Accuracy, affordability, timeliness and accessibility are critical factors that impact diagnosis quality. To improve diagnosis accuracy, affordability, accessibility and eliminate biases, machine learning and deep learning-based algorithms are gaining attraction in revolutionizing the healthcare industry. The objective of this study is to create a flexible and comprehensive medical diagnostic framework based on machine learning and deep learning models that can predict many diseases based on a patient's health records. The main objective of this study is to prevent problems caused by misdiagnosis and delayed diagnosis. By analyzing multiple diseases using a single platform, the cost of patient treatment can be reduced significantly, making it more accessible for people in underprivileged regions. By increasing the accuracy and speed of disease prediction, the proposed machine learning and deep learning-based diagnosis system has the potential to save lives. Experimental results showed that RandomForest has outperformed competing models on numerical datasets while on image datasets, VGG16 generated best accuracy than ResNet50.

Objective: To determine the mean C-reactive protein (CRP) levels preoperatively among patient with Acute Cholecystitis to predict difficulty of Laparoscopic Cholecystectomy on Nassar Grade. Methodology: This cross-sectional study was conducted in General Surgery Department, PAEC General Hospital, Islamabad from 30/01/2023 to 31/07/2023 This study was conducted in the Department of General Surgery, PAEC General Hospital, Islamabad. 150 patients having symptomatic gallstones were enrolled in the study. CRP level was determined preoperatively in all patients. All patients underwent LC. Intraoperative difficulty level was determined using the Nassar grade. Mean CRP level was compared between the 5 grades Nasser scale, by applying ANOVA test keeping p-value ? 0.05 as significant. Results: A total of 150 patients with symptomatic gallstones were enrolled for the study. The mean age of the patients was 47.29+/-13.75 years. 34 (22.7%) were male and 116 (77.3%) were female. The mean WBC count of our patients was 9075.27+/-3437.46/µL. The mean CRP level of our patients was 31.01+/-61.32 mg/dL. The mean CRP level was significantly higher in patients with grade 4, followed by grade 3, grade 2 and grade 1 (p-value < 0.01). Conclusion: Our study concluded that there is a significant link between higher CRP levels and increased Nassar grades, indicating a positive relationship between elevated CRP levels and greater intraoperative difficulty during LC.

Human Resource Management Systems (HRMS) have become indispensable for modern organizations aiming to streamline workforce operations and optimize resource allocation. By integrating employee data into a centralized platform, HRMS enhances transparency, ensures compliance, and supports evidence-based decision-making. In the context of the Khyber Pakhtunkhwa (KP) government, the implementation of HRMS has revolutionized the management of public sector employees, particularly within the healthcare domain. This initiative addresses challenges such as data discrepancies, resource inefficiencies, and policy compliance gaps, ensuring a more efficient and accountable workforce. My tenure as a Monitoring & Validation Officer at Palladium Group, Islamabad, between June 2024 and December 2024, provided first-hand insights into the transformative potential of HRMS in improving workforce management practices

Introduction: High-dose melphalan is the standard conditioning chemotherapy for autologous stem cell transplantation (ASCT) in multiple myeloma (MM). However, patients experience several side effects and toxicities from high-dose melphalan. In 2016, United States Food and Drug Administration approved Evomela, a propylene glycol-free formulation of melphalan, as conditioning chemotherapy for ASCT in MM. This was based on its bioequivalence to the standard propylene-glycol solubilized melphalan formulation (Alkeran) in a phase 2 study. Evomela has the advantages of improved solubility, stability, bioavailability and being free of propylene glycol that is associated with organ dysfunction. Methods: We conducted a retrospective study of patients who received ASCT with high dose chemotherapy using alkeran (n=255) or evomela (n=259) at our institution to compare their outcomes such as side effects, duration of cytopenias, transfusion requirements, length of hospital stay, readmission within 30 days and progression-free survival (PFS) post-SCT. Clinical and demographic characteristics were compared between two treatment regimens using the Chi-square test for categorical variables and the Wilcoxon rank sum test for the continuous variables. Overall survival (OS) was calculated from the date of transplant to death, censoring the alive patients at their last follow up date. Progression-free survival (PFS) was calculated from the date of transplant to date of relapse or death, whichever occurred first, censoring at the last follow-up if no relapse or death. OS and PFS estimates were calculated using the Kaplan-Meier method and compared using the log-rank test. Results: The baseline patient characteristics such as age, ISS stage, comorbidity index and number of prior lines of therapy prior to ASCT were similar between the two groups. (See table 1). Mucositis was seen in 77.2% of the patients who received Alkeran compared to 69.5% who received Evomela (p=<0.001). Incidence of febrile neutropenia was 65.9% in the Alkeran group and 49.4% in the Evomela group (p=0.0002). Chemotherapy-induced nausea and vomiting were reported in 98.8% and 93.4% of the patients in the Alkeran and Evomela groups respectively (p=0.001). Rates of diarrhea and clostridium difficile infection were similar with the two drugs. Time to neutrophil engraftment was the same in both the groups while duration of thrombocytopenia (platelets <20k) was slightly longer in the Evomela group (6 days in alkeran and 8 days in evomela group, p=<0.001). Red cell transfusion requirement was higher with the use of Alkeran compared to Evomela (42.3% vs 21.8%, p=0.001) while platelet transfusion was the same. There was no difference in the duration of hospital stay between the two groups. However, rate of readmission within 30 days of discharge was higher in patients who got Evomela compared to Alkeran (9.4% versus 17.4%, p=0.008). Day +100 serological response (very good partial response or better), PFS post-SCT and OS were similar in both groups. (Figure 1). Conclusion: We conclude that use of Evomela is associated with a better side-effect profile and transfusion requirement while having similar outcomes as Alkeran. Disclosures Yucebay: Janssen: Membership on an entity's Board of Directors or advisory committees; BioXCell: Membership on an entity's Board of Directors or advisory committees; G1 Therapeutics: Membership on an entity's Board of Directors or advisory committees. Chaudhry:Sanofi: Consultancy, Membership on an entity's Board of Directors or advisory committees. Efebera:Takeda: Honoraria, Speakers Bureau; Pharmacyclics: Research Funding; Celgene: Research Funding; Ohio State University: Current Employment. Bumma:Amgen: Speakers Bureau; Sanofi: Speakers Bureau. Khan:Amgen: Consultancy; Janssen: Consultancy. Devarakonda:Janssen: Consultancy.

We propose to perform an image-based framework for electrical energy meter reading. Our aim is to extract the image region that depicts the digits and then recognize them to record the consumed units. Combining the readings of serial numbers and energy meter units, an automatic billing system using the Internet of Things and a graphical user interface is deployable in a real-time setup. However, such region extraction and character recognition become challenging due to image variations caused by several factors such as partial occlusion due to dust on the meter display, orientation and scale variations caused by camera positioning, and non-uniform illumination caused by shades. To this end, our work evaluates and compares the stateof-the art deep learning algorithm You Only Look Once (YOLO ) along with traditional handcrafted features for text extraction and recognition. Our image dataset contains 10,000 images of electrical energy meters and is further expanded by data augmentation such as in-plane rotation and scaling to make the deep learning algorithms robust to these image variations. For training and evaluation, the image dataset is annotated to produce the ground truth of all the images. Consequently, YOLO achieves superior performance over the traditional handcrafted features with an average recognition rate of 98% for all the digits. It proves to be robust against the mentioned image variations compared with the traditional handcrafted features. Our proposed method can be highly instrumental in reducing the time and effort involved in the current meter reading, where workers visit door to door, take images of meters and manually extract readings from these images.