Jaramogi Oginga Odinga University of Science and Technology

AfriVoices-KE is a large-scale multilingual speech dataset comprising approximately 3,000 hours of audio across five Kenyan languages: Dholuo, Kikuyu, Kalenjin, Maasai, and Somali. The dataset includes 750 hours of scripted speech and 2,250 hours of spontaneous speech, collected from 4,777 native speakers across diverse regions and demographics. This work addresses the critical underrepresentation of African languages in speech technology by providing a high-quality, linguistically diverse resource. Data collection followed a dual methodology: scripted recordings drew from compiled text corpora, translations, and domain-specific generated sentences spanning eleven domains relevant to the Kenyan context, while unscripted speech was elicited through textual and image prompts to capture natural linguistic variation and dialectal nuances. A customized mobile application enabled contributors to record using smartphones. Quality assurance operated at multiple layers, encompassing automated signal-to-noise ratio validation prior to recording and human review for content accuracy. Though the project encountered challenges common to low-resource settings, including unreliable infrastructure, device compatibility issues, and community trust barriers, these were mitigated through local mobilizers, stakeholder partnerships, and adaptive training protocols. AfriVoices-KE provides a foundational resource for developing inclusive automatic speech recognition and text-to-speech systems, while advancing the digital preservation of Kenya's linguistic heritage.

Tuberculosis (TB) is a top global health challenge, with 1.2 million children aged 0-14 years becoming ill with TB every year. Yet, a significant proportion remain undiagnosed or missed due to diagnostic barriers. This cross-sectional study employing an embedded mixed method approach investigated factors contributing to low tuberculosis diagnosis among children in Gem Sub County, Siaya County Kenya, a high burden region. Data was collected from 71 healthcare workers (HCWs) and 16 respondents across six wards using semi-structured questionnaires and TB register abstraction. Qualitative data underwent deductive thematic analysis while quantitative data was analyzed using descriptive statistics and logistic regression in SPSS version 27. About a third (31.3%) of TB cases in children required three or more facility visits before diagnosis, with some needing up to 12 visits. Costly and inaccessible chest X-rays and GeneXpert underutilization due to difficulties in sputum and alternative sample (gastric lavage, nasopharyngeal aspirates) collection were key diagnostic constraints. Delayed care-seeking due to stigma, misconceptions (56.3% linked childhood TB to HIV) and low TB symptom awareness (37.5% of children presented with ≤2 symptoms) were patient related factors associated with low TB diagnosis. About 56.3% of caregivers sought routine TB screening, but 62.5% sought care only after persistent symptoms. Clinical officers showed substantially higher odds of reporting confidence in sample collection (OR=34, 95% CI 3.81-303.21, p = 0.002) and GeneXpert interpretation (OR=15, 95% CI 3.20-70.39, p < 0.001) than nurses. Trained HCWs were 7.16 times more likely to interpret X-rays accurately (95% CI 2.16-23.67, p < 0.001). Enhanced HCW training, improved diagnostic access, targeted community education on stigma and misconceptions are critical for early TB detection in children.

ABSTRACT Small‐scale fisheries and aquaculture constitute critical pillars of food security, livelihoods, and rural economies across East Africa, yet persistent gendered inequalities continue to constrain both equity and sectoral performance. This study provides a synthesis of empirical evidence from Kenya, Uganda, and Tanzania to examine how structural, institutional, and socio‐cultural processes reproduce gender disparities across fisheries and aquaculture value chains. Drawing on a narrative review of peer reviewed and grey literature published between 2015 and 2025, the analysis applies the Harvard Analytical Framework (HAF) to interrogate four interrelated dimensions: Division of labour, access to and control over productive resources, influencing institutional and normative factors, and gender integration within project cycles. The findings indicate that gender inequalities are most pronounced in asset ownership, decision making authority, and benefit distribution rather than participation alone. Women remain concentrated in labour intensive and lower value post‐harvest activities, while men dominate harvesting, aquaculture production, licensing systems, and leadership positions within co management institutions. Across the three countries, men retain control over the majority of productive assets, including boats, fishing gear, land, and aquaculture infrastructure, while women's access to credit, extension services, and modern technologies remains constrained. These disparities are reinforced by inheritance regimes, gender norms, and male dominated governance structures that shape access to resources and control over income. Development interventions frequently expand women's participation without addressing underlying structural constraints, thereby contributing to the reproduction of institutional inequalities. By linking structural analysis to policy relevant pathways, the study argues that inclusive blue economy development requires tenure reform, gender responsive financing mechanisms, institutional restructuring, and sustained normative change. The review positions gender equity as a governance and productivity imperative and provides an evidence based foundation for advancing resilient and inclusive fisheries and aquaculture systems in East Africa.

This study reports large microplastics found in Lake Victoria through an analytical workflow that combines the complementary methods stereomicroscopy, micro-Fourier transform infrared (µ-FTIR) spectroscopy and pyrolysis gas chromatography-mass spectrometry (Pyr-GC-MS) for the quantification of 11 environmentally relevant microplastic polymers. Algae-rich surface water samples (n = 18) were trawled using a 0.3 mm manta net from Lake Victoria, the world’s largest tropical lake. Stereomicroscopy as a determinant analytical technique detected 191 particles, which were primarily blue fragments and fibres. Polyethylene and polypropylene were the dominant polymers identified by µ-FTIR. Pyr-GC-MS allowed the detection and quantification of microplastics (MPs) with LOD and LOQ of 0.01–14.7 µg and 0.03–49.1 µg. Polyethylene (0.058–0.34 µg/L), polypropylene (0.024 µg/L and 0.043 µg/L), nylon 6 (0.0051–0.064 µg/L), nylon 66 (0.0022–0.084 µg/L), poly(ethylene terephthalate) (0.0029–0.027 µg/L) and poly(methyl methacrylate) (0.0036 µg/L) were quantified. µ-FTIR was found to be suitable for the identification of the most abundant polymers in the 0.3–4.9 mm size range whereas Pyr-GC-MS afforded the quantification of seven polymers, most of which were not detected by µ-FTIR. This complementary workflow gave a wider perspective on MP loading, providing both polymer concentrations and physical characteristics (sizes, colours, forms and count) of the MPs.

Arrhythmia is a condition in which a person's heartbeat is irregular and can pose serious health risks. Effective arrhythmia detection is necessary to reduce associated risks. This study aims to develop a new deep learning architecture combining Convolutional Neural Networks (CNNs) to extract features, Long Short-Term Memory (LSTMs) to handle sequential data, and Gated Recurrent Units (GRUs) to reduce computational resources, leveraging the strengths of each to achieve better classification accuracy for diagnosing using the MIT-BIH Arrhythmia Database. The data is preprocessed by 0.5 Hz (low) and 50 Hz (high) to remove noise, then segmented into smaller, normalized signals to a unique scale, determine the higher point of the QRS complex in ECG, then gets labeled each segment, and finally converted ECG segment to (2D). The proposed model outperforms models on CNN, LSTM, and GRU if we apply them alone, with a precision of 92%, F1-score of 97%, recall of 100%, and accuracy of 97%; this study's notable discovery is that the suggested method may substantially decrease the duration when using RNN networks in conjunction with CNN. This paper presents a cost-effective approach to ECG signal reduction and a robust automatic scheme for arrhythmia detection, leveraging the strengths of CNN, LSTM, and GRU networks. The suggested model has achieved significant improvements in accuracy and is potentially a useful tool for real-time clinical practice.