Journal of Computational Science and Data Analytics

Profiling of Breast Cancer Prevalence and Its Diagnosis Using Varied Imaging Techniques in Tanzania

2025-02-12T18:22:21+00:00

Breast cancer is not only the most commonly occurring cancer among women, but also the most frequent cause of cancer-related deaths in women in developing countries. The mortality rate is marginally higher in developing countries than in developed countries with about 60% of the deaths occurring in developing countries. In Tanzania for exam ple, breast cancer is the second leading cancer in terms of incidence and mortality among women after cervical cancer. Approximately half of all women diagnosed with breast can cer in Tanzania die of the disease. This is due to the limited number of medical facilities for cancer screening and diagnosis, the limited number of oncologists and pathologists, and the diagnosis costs in the country. Due to the mentioned factors, it is approximated that, 80% of breast cancer cases in Tanzania are diagnosed at advanced stages (III or IV), when treatment is less effective and outcomes are poor. By 2030, new breast can cer cases are approximated to increase by 82% in Tanzania. The diagnosis/screening of breast cancer starts with breast imaging with ultrasound and mammograms. Suspected cases are then subjected to pathology for confirmatory tests. Although breast imaging plays a major role in both breast cancer screening and diagnosis, the service is largely not available in many developing countries. Our study found the absence of routine breast cancer screening in Tanzania, resulting in late-stage detection of many cases. This is mainly due to a lack of enough well-trained radiologists to read the images and the costs of the process. This study is aimed at exploring the role, importance and challenges of breast imaging in the screening and diagnosis of breast cancer in Tanzania, a developing country. It is worth noting that, breast imaging is an important step in screening for breast cancer. Our results found that, there is a significant number of malignancies under the recommended age of breast cancer screening of fifty years of age. Our study also found a very high Inter variability among radiologists. This study also discovered in our sample size that 66% of patients did not have their samples taken for confirmation by the pathologists. This might be due to the costs of the process or loss of follow-ups as many patients came far from the diagnosis Centre. Due to the higher intervariability among radiologists, this suggests the necessity of at least two radiologists reading the same case before the conclusion of the diagnosis. Also, due to the significant number of malignancies under the recommended age of 50 years, this study recommends the age to be reconsidered based on different settings. Due to the challenges observed in breast imaging, this study recommends the use of computer-aided diagnosis (CAD) with Artificial Intelligence to assist the limited number of radiologists available.

Improved Pilot Decontamination Scheme for Massive MIMO Networks

2025-08-11T06:18:51+00:00

Improved pilot reuse and channel estimation approaches are proposed in this study to mitigate pilot contamination in massive Multiple input multiple Input (MIMO) net works. Pilot contamination is a major issue in circumstances with a high user density, where interference from adjacent users limits system efficiency. We propose a solution to address this by classifying users into active and idle groups based on their resource needs and channel circumstances, hence prioritizing active users for pilot assignment. A weighted graph coloring approach is used to assign pilot symbols to active users, minimizing interference by allocating neighboring users to distinct pilots. Based on their channel gain, each user’s weight indicates their respective signal strength. In order to improve the quality of channel state information and further minimize contamination, we further investigated channel estimation techniques. The findings of the simulation demonstrate that compared to the weighted graph coloring currently in use, the improved pilot contamination mitigation technique offers notable improvements in spectrum efficiency. In dense network environments, these results show how the suggested methodology may enhance the performance of MIMO systems, offering a workable method for system optimization and resource allocation.

Deep Reinforcement Learning Based Synergies Pushing and Grasping Policies in Cluttered Scene Using UR5 Robot

2024-12-20T06:35:02+00:00

This paper presents deep reinforcement learning-based synergy between pushing and grasping systems to improve the grasping performance of the UR5 robot in a cluttered scene. In robotic manipulations, grasping an object in a clutter is fundamental yet a challenging activity for industrial applications. This is because numerous studies focused on improving grasping performance in cluttered environments using either a grasping-only policy or pushing and grasping without incorporating a pushing reward. Additionally, some research has been limited to using similar objects, such as cubes. This paper for mulated the mathematical modeling of the universal robot manipulator. The proposed model involves training two fully-connected convolutional neural networks that transfer visual observations of the scene to a dense pixel-wise sample of end-effector orientations and positions for each pushing and grasping action. A fixed RGB-D camera is used to take the raw images of the scene and generate a heightmap image. Before feeding the heightmap image to the fully convolutional network, it is rotated by 36 different angles to generate 36 pixel-wise Q-value predictions. Both pushing and grasping networks are self-supervised by trial and error from experience and are trained together in a deep Q-learning algorithm. Successful grasps have a reward of 1, while successful pushes have a 0.5 reward value. But unsuccessful actions have a reward of 0 value. The proposed policy learns pushing motions to improve future grasping in a cluttered scene. The experiment demonstrates that the proposed model can successfully grasp objects with an 87% grasp success rate while grasping only policy, no-reward for pushing policy, and stochastic gradient without momentum is 60%, 71%, and 79% respectively. Further, it has been demonstrated that the proposed model is capable of generalizing to randomly arranged cluttered objects, challenging arrangements, and novel objects.

Using Public Health datasets to predict one’s ability to pay for Pre-Exposure prophylaxis (PrEP) services in Uganda

2025-09-02T09:49:44+00:00

In Uganda, the uptake of pre-exposure prophylaxis (PrEP) as a preventive measure against HIV infection is notably low, despite its proven effectiveness, particularly among high-risk populations (UPHIA, 2020). Although PrEP has historically been available at no cost in government facilities, the recent decrease in HIV medication costs and the shift towards private-sector involvement necessitate a reliable assessment of individuals’ ability to pay for PrEP. The growing volume of HIV-related data presents a unique opportunity to leverage artificial intelligence (AI) and machine learning (ML) techniques to identify high-risk sub-populations that are both eligible for and willing to pay for PrEP services. This retrospective study, analyzed three diverse datasets, including, the Uganda Demographic Health Survey, the Uganda Population HIV/AIDS Impact Assessment survey, and a private dataset from the Rocket Health Telemedicine Clinic. The study population included individuals aged 18 years and above that have accessed a private health facility for sexual reproductive health services or products. Statistical methods, including the Chi-square test and Spearman’s correlation test, were employed to identify features with a statistical significance to the ability to pay for PrEP. The datasets were aggregated, cleaned and then split into 70% for training and 30% for testing and validation. An ensemble of machine learning classification models was trained using Python and the PyCaret library. The AdaBoost classifier demonstrated superior predictive power, with a recall of 99% and an AUC of 100%, indicating robust prediction capabilities on this dataset. The model achieved a high training score of 99%, suggesting an excellent fit to the training data. Further analysis revealed that factors such as age, gender, employment status, and socioeconomic status were the most influential predictors of the ability to pay for PrEP services. A web application interface was developed using the Streamlit library, allowing individuals and programs to upload data and make predictions about the likelihood of individuals paying for PrEP. The developed tool leverages publicly available data to identify populations capable of paying for PrEP services, fostering a collaborative effort towards achieving better health outcomes and ensuring the sustainability of HIV prevention services.

BehFayda: A Comprehensive Review and Framework Proposal for Adaptive Authentication in National Identity Systems Using Multi-Modal Biometric Fusion

2025-01-01T07:35:27+00:00

The proliferation of digital services necessitates robust identity verification mechanisms. The Ethiopian digital national ID, Fayda, built on the Modular Open-Source Identity Platform (MOSIP), aims to offer a secure and scalable solution for national identity management. However, MOSIP lacks explicit support for adaptive continuous authentication—a crucial aspect of ensuring security and usability. This paper introduces BehFayda, a comprehensive architecture for a privacy-enhanced multi-modal biometric fusion system for adaptive continuous authentication tailored to digital identity systems. The framework integrates behavioral biometrics, such as keystroke dynamics in two languages, swipe gestures, motion data, and contextual data as a candidate for the proposed fusion strategy. We propose the Multi-Modal Deep Residual Fusion (MM-DRF) algorithm, which incorporates feature-level fusion with adaptive attention mechanisms to dynamically adjust the contribution of different biometric modalities based on their relevance. Our approach provides a new insight to enhance authentication accuracy which mainly aims to guide future research in advancing adaptive authentication in national digital identity systems, with a focus on privacy-preserving techniques and real-time behavioral analysis.