http://localhost:8080/xmlui/handle/123456789/20 Wed, 12 Aug 2026 15:38:34 GMT 2026-08-12T15:38:34Z http://localhost:8080/xmlui/handle/123456789/12817 A Fairness-Aware Framework for Bias Mitigation in Diagnostic Skin Imaging Using an Adversarial CNN Approach Waithira, Joshua Mwaura The integration of Artificial Intelligence (AI) into dermatological diagnostics offers great potential for improving clinical decision- making and patient outcomes. However, persistent biases in training data and model structures continue to increase disparities in diagnostic accuracy, especially across different skin tones and lesion types. These inequalities emphasize the urgent need for fairness - aware methods that provide equitable performance in clinical settings. Therefore, this study aimed to develop and validate a fairness - aware AI framework for dermatological imaging, designed to reduce bias while maintaining diagnostic accuracy. The framework was guided by three goals: (i) to assess how well adversarial debiasing techniques can remove bias- related features, (ii) to analyze current diagnostic systems and identify fairness gaps, and (iii) to introduce a customized bias mitigation strategy for skin lesion classification. Adversarial debiasing was applied through a Gradient Reversal Layer (GRL) within a Convolutional Neural Network (CNN). The GRL works by reversing gradient signals during training, discouraging the network from encoding bias- related attributes while still optimizing for diagnosis accuracy. This mechanism formed the core of the fairness - aware framework. Performance was evaluated by comparing the framework to a baseline CNN using the ISIC 2020 skin lesion dataset, which includes 33, 126 dermoscopic images from over 2, 000 patients. Preprocessing and augmentation techniques were used to improve data quality and model robustness. External validation utilized the Fitzpatrick 17 k dataset, containing 16, 577 clinical images annotated by skin type, to test demographic generalizability. Expert validation was also performed, with dermatologists reviewing interpretability outputs to confirm clinical relevance and practical use. Model interpretability was improved through SHAP- based feature attribution, helping clinicians visualize decision boundaries and better understand the framework's diagnostic reasoning. The fairness- aware framework significantly reduced the Statistical Parity Difference from. .35 to. .05, while maintaining high diagnostic accuracy, sensitivity, and AUC--ROC scores above. .87. Fairness improvements were supported by Equalized Odds metrics, and statistical testing showed that these equity gains did not compromise reliability. These results highlight the clinical potential of fairness- aware AI in dermatology. By combining adversarial debiasing, expert validation, and practical application, the proposed framework offers actionable paths for the ethical use of AI in precision medicine. More broadly, it advances the discussion on bias mitigation in healthcare AI, demonstrating the transformative potential of equity- focused approaches to improve diagnosis outcomes across diverse populations. Thu, 16 Apr 2026 00:00:00 GMT http://localhost:8080/xmlui/handle/123456789/12817 2026-04-16T00:00:00Z http://localhost:8080/xmlui/handle/123456789/12814 Influence of Records Management Policies on Service Delivery in County Governments: A Case of County Government of Isiolo Guyo, Tacho Mohamed The “purpose of this study was to investigate the influence of records management policies on service delivery in County Governments, a case of County Government of Isiolo. Records management policy is essential to records management and helps protect the organization from the negative effects of poor service delivery. The study was carried out at the county government of Isiolo headquarters and included respondents from departments of records registry, human resources, finance, audit, supply chain management and ICT. The specific objectives were to: Determine the influence of records creation and capture policy on service delivery; assess the influence of records storage and security policy on service delivery; Evaluate the influence of records retention and disposal policy on service delivery, and propose a records management framework. The study used a descriptive research design and a mixed-method approach incorporating qualitative and quantitative techniques. Both conceptual and empirical literature were reviewed based on findings and comments on existing research, publications and journals by other scholars. Data collection was necessitated by obtaining research authorization letters from relevant institutions. Structured questionnaires and interview schedules were pretested to determine validity and reliability and thereafter employed to collect data. Kaiser-Meyer-Olkin Measure of Sampling Adequacy for records creation and capture policy, records storage and security policy, records retention and disposal policy, records management framework and service delivery were 0.634, 0.671, 0.783, 0.735 and 0.652, respectively, thus, construct validity was achieved. On the other hand, Cronbach's alpha value of 0.753, 0.758, 0.766, 0.708, and 0.702 was established for Records creation and capture policy, records storage and security policy, records retention and disposal policy, records management framework and service delivery, respectively. The target population comprised 242 employees at the county headquarters and distributed among the six departments that deal with records directly or indirectly. Stratified sampling technique was used to classify employees into strata. Yamane's simplified formula was used to obtain a sample size of 151 respondents proportionately distributed per department using a random sampling technique. Quantitative data was analyzed using descriptive statistics with the aid of IBM–SPSS (Version 23), presented in a table and interpreted in mean, frequencies, percentages, and standard deviation. Qualitative data was analyzed using content analysis. Ordinal regression analysis was used to estimate the model. Comparative analysis methods like generalized linear models were further used to investigate the relationship between service delivery and Records Management Policies. The results of this study will assist county government policymakers, records professionals, and researchers by giving insights into the implementation of record management policies in influencing service delivery. The findings revealed an Exp(B), which is the antilog of a regression coefficient (B) in parameter estimate table showed how a change in a predictor affects the outcome in an understandable way. The outcome of 1.912, 2.105, 1.818 and 1.820 for records creation and capture policy, records storage and security policy, records retention and disposal policy, and records management framework, respectively were positively realized, suggesting an increasing probability of being in a higher level of service delivery as values on record creation and capture policy, records storage and security policy, records retention and disposal policy, and records management framework increase. The study concluded that records management policies influenced service delivery in the County Government of Isiolo. The study also recommends adequate consideration and full implementation of records creation and capture policy, records storage and security policy, records retention and disposal policy and records management framework by the County Government of Isiolo to ensure enhanced service delivery. This research is expected to strengthen records management in Isiolo County through adoption and implementation of the proposed framework, thereby improving service delivery.” Wed, 06 May 2026 00:00:00 GMT http://localhost:8080/xmlui/handle/123456789/12814 2026-05-06T00:00:00Z http://localhost:8080/xmlui/handle/123456789/10122 Designing a Machine Learning Model for Crop Pest Surveillance among Small Holder Farmers in Uasin Gishu County, Kenya Songol, Michael Kipkorir Uasin Gishu County is noted as a breadbasket region in Kenya. This county provides a big proportion of the rural smallholder farmers’ income because 90% of its land is arable. It is endowed with high and consistent rainfall and favorable cropping temperatures. Maize, wheat, beans, and Irish potatoes are the common food crops in the county. However, crop productivity in the region is currently tapering off due to the use of traditional mechanisms to mitigate and control emerging crop pests and their effects. This coupled with inadequate extension services and hardly accessible information from agricultural agencies both at the county and the national governments has greatly contributed to a decline in agricultural production in the county. Crop pest surveillance by small holder farmers has been sub optimal while using ICT tools such as mobile phone. Notably, this challenge can be mitigated by leveraging mobile technology in building digital solutions that can provide farmers with easily accessible, precise, and timely information. Other technologies used in modern world applications include Machine Learning (ML), a branch of artificial intelligence. The digital solution in this case is ML technology which has proved to minimize losses incurred in farming. In solving the real-world problems, ML has found its use in predicting commodity prices, detecting fraudulent transactions, treatment and diagnosis of diseases, prediction of energy use and image recognition among other uses. It has increased efficiency and precision in farming thereby guaranteeing high quality farm output. The study aimed at exploring potential of ICT tools in providing information access to farmers by leveraging mobile technology. The main objectives looked at establishing usage of mobile phones for crop pest surveillance in Kenya, to design a machine learning model for crop pest surveillance for small holder farmers, and to evaluate the proposed machine learning model for crop pest surveillance among small holder farmers using mobile phones. The study targeted farmers in Kesses Sub County involved in small scale crop farming. Stratified sampling technique was used to select 3 Wards in Kesses Sub County that have had the highest hit of emerging crop pests. The targeted sampling technique was used to select farmers who experienced greatest loss due to crop pest between 2017 and 2021. The study involved mixed methods research design, and system analysis design methodologies. In mixed methods design, questionnaires were administered to respondents. Statistical methods employed was descriptive statistics and correlation analysis. In analyzing qualitative data, content analysis technique was applied. In systems analysis and design, this described how the development of the system will be achieved using the four-phased model which are planning, analysis, design, and implementation. Insights from field data showed that most farmers, 45 percent are using mobile phones in sourcing agriculture information. This was used to inform design of machine learning model for crop pest surveillance. This was achieved by use of plant disease detection, processing, segmentation, extraction, and classifier algorithms. In supporting the testing of the algorithm, accuracy, precision, recall, and F1 measurement score was used because this is highly supported in literature results with over 90% scores, which is within the acceptable score for testing. This study supports the use of mobile phone as one of key tools in carrying out crop pest surveillance by small holder farmers. Wed, 01 Oct 2025 00:00:00 GMT http://localhost:8080/xmlui/handle/123456789/10122 2025-10-01T00:00:00Z http://localhost:8080/xmlui/handle/123456789/9967 A Cybersecurity Defense Framework for Healthcare: A Case of Moi Teaching and Referral Hospital, Eldoret Bore, K. Jacob In the face of escalating cybersecurity threats, tailored defense frameworks are imperative to safeguard patient data and ensure the continuity of critical healthcare services. This study pursues four primary objectives: (I) to identify and analyze evolving cybersecurity threats and challenges in the healthcare sector, (II) to establish cybersecurity strategies and countermeasures for healthcare organization systems, (III) to design and develop a cybersecurity framework specifically tailored for the healthcare sector, and (IV) to validate the cybersecurity framework within the healthcare sector. A comprehensive approach was adopted, commencing with a meticulous literature review of existing cybersecurity frameworks. Subsequently, qualitative research was conducted through interviews in key departments at Moi Teaching and Referral Hospital (MTRH), including management of HRIS, ICT, and Internal Audit departments. These departments serve as primary custodians of data governance systems, tools, and policies, providing insights into cybersecurity practices within the healthcare sector, elucidating both strengths and gaps. Identified gaps include inconsistencies in data access controls, insufficient real-time monitoring capabilities, and a lack of automated compliance monitoring mechanisms. Building upon these findings, a novel cybersecurity defense framework was meticulously crafted, incorporating elements from HL7 and the Kenya Health Act. Validation of the framework was conducted through a mixed-method approach, including questionnaires and interview schedules at MTRH. The positive responses obtained affirm the efficacy of the framework, underscoring its relevance in fortifying cybersecurity practices within the healthcare sector. Following the validation of the framework, recommendations were derived. These recommendations include the implementation of fine-grained access controls to mitigate unauthorized data access, the enhancement of real-time monitoring systems to promptly identify and respond to cybersecurity incidents, and the integration of automated compliance monitoring tools to ensure adherence to industry regulations. Additionally, the study recommends ongoing training and awareness programs for healthcare staff to bolster cybersecurity awareness and adherence to best practices. Overall, this thesis contributes to advancing cybersecurity practices, ensuring secure and uninterrupted delivery of high-quality patient care amidst a hostile digital landscape. Wed, 01 Oct 2025 00:00:00 GMT http://localhost:8080/xmlui/handle/123456789/9967 2025-10-01T00:00:00Z