ABSTRACT
MRI and CT are the two most widely used medical imaging techniques. Often, doctors need images from both modalities to diagnose conditions accurately and plan treatments like radiation therapy. However, using both MRI and CT can be costly, and it often results in misaligned images. A practical alternative is to use computational methods to convert images from one modality to another—particularly converting MRI images into CT images. In this study, we explore a deep learning approach using diffusion models and score-matching techniques to address this challenge. Specifically, we adapt denoising diffusion probabilistic models and score-matching strategies, apply four different sampling methods, and compare their performance with that of traditional models like generative adversarial networks (GANs) and convolutional neural networks (CNNs). Our results show that diffusion and score-matching models generate synthetic CT images with higher quality than CNN and GAN approaches. We also assess the uncertainty in these models using Monte Carlo simulations and further improve the final image quality by averaging the Monte Carlo outputs. Overall, our research suggests that diffusion and score-matching models not only rival CNNs and GANs in generating cross-modality medical images but also offer a more mathematically grounded and reliable framework.

ABSTRACT
MRI and CT are the two most widely used medical imaging techniques. Often, doctors need images from both modalities to diagnose conditions accurately and plan treatments like radiation therapy. However, using both MRI and CT can be costly, and it often results in misaligned images. A practical alternative is to use computational methods to convert images from one modality to another—particularly converting MRI images into CT images. In this study, we explore a deep learning approach using diffusion models and score-matching techniques to address this challenge. Specifically, we adapt denoising diffusion probabilistic models and score-matching strategies, apply four different sampling methods, and compare their performance with that of traditional models like generative adversarial networks (GANs) and convolutional neural networks (CNNs). Our results show that diffusion and score-matching models generate synthetic CT images with higher quality than CNN and GAN approaches. We also assess the uncertainty in these models using Monte Carlo simulations and further improve the final image quality by averaging the Monte Carlo outputs. Overall, our research suggests that diffusion and score-matching models not only rival CNNs and GANs in generating cross-modality medical images but also offer a more mathematically grounded and reliable framework.

ABSTRACT
Introduction: Paediatric Computed Tomography (CT) scans are essential diagnostic tools, but they raise important concerns due to children's increased vulnerability to radiation risks compared to adults. In Zambia, the lack of locally established paediatric Diagnostic Reference Levels (DRLs) hampers efforts to standardise protocols and optimise radiation doses. This study introduces the first local DRLs and estimates the effective doses for paediatric head CT scans in Lusaka Province. Methods: A retrospective, cross-sectional study was conducted at three public hospitals in Lusaka, involving data from 337 paediatric patients aged 0-18 years who underwent head CT scans between 2019 and 2023. Effective Doses (ED) were calculated using age-specific conversion factors based on age-based DRLs, according to the ICRP protocol. Results: The established local DRLs showed an age-related increase, indicating the need for higher radiation parameters in older children. Median (IQR) DLP values ranged from 551.5 mGy·cm (IQR 456.10-653.80) for the <1-year age group to 793.4 mGy·cm (IQR 701.00-1210.00) for the 10-15-year age group. The overall median effective dose was 4.66 mSv (IQR 2.605-7.375 mSv). Significant differences in DLP values between centres were observed (p<0.0001). Patient age showed a moderate positive correlation with both CTDIvol (r=0.375, p<0.0001) and DLP (r=0.476, p<0.0001). Compared to international benchmarks, the DRLs displayed notable differences. Conclusion: This study has established local Diagnostic Reference Levels (DRLs) for paediatric head CT scans in Lusaka, Zambia. Variations in CT doses were observed across hospitals. It is advised that CT imaging protocols for paediatric patients be optimised. Collaboration between radiographers, radiologists, and medical physicists, along with comprehensive training for CT radiographers, is also recommended.

ABSTRACT
Hepatocellular carcinoma (HCC) is among the leading causes of cancer-related deaths worldwide. Transarterial chemoembolization (TACE) remains the standard locoregional therapy for intermediate-stage disease. Traditionally performed via the transfemoral approach (TFA), the transradial approach (TRA) has emerged as a promising alternative due to fewer access site complications, shorter hemostasis time, and greater patient comfort. This comparative observational study, conducted at a tertiary care hospital, evaluated 38 HCC patients undergoing TACE—19 via TRA and 19 via TFA. Technical success, fluoroscopy time, contrast volume, post-procedural pain, hemostasis time, hospital stay, and cost were analyzed. Technical success was comparable between TRA (94.7%) and TFA (89.5%). However, TRA showed significantly reduced hemostasis time (15.8 ± 3.4 min vs. 22.1 ± 4.2 min, p = 0.001), lower pain scores (2.5 ± 1.3 vs. 3.9 ± 1.7, p = 0.019), shorter hospital stay (12.5 ± 2.1 h vs. 16.8 ± 3.5 h, p = 0.008), and lesser contrast use (65.2 ± 8.1 mL vs. 72.5 ± 9.3 mL, p = 0.021). Access site complications were lower with TRA (10.5% vs. 15.8%, p = 0.559) though not statistically significant. Overall, TRA demonstrated improved procedural efficiency, patient comfort, and cost-effectiveness without compromising technical success. These findings reinforce the safety and practicality of the transradial route for hepatic arterial embolization in HCC patients undergoing TACE. Approach depends largely on operator expertise and familiarity with radial artery anatomy, as emphasized by Al-Hakim et al., (2016) in Cardiovascular and Interventional Radiology. The study also corroborated that socio-demographic variables, such as age, gender, BMI, smoking status, alcohol consumption, and comorbidities, did not significantly impact procedural outcomes, indicating that transradial access is suitable for a diverse population.

ABSTRACT
The Prostate Imaging Reporting and Data System (PIRADS) is a cornerstone of modern prostate cancer diagnostics, standardizing the acquisition, interpretation, and reporting of multiparametric Magnetic Resonance Imaging (mpMRI). This article provides a comprehensive guide to PIRADS version 2.1, detailing its structured algorithm for assigning a risk score from 1 to 5 that correlates with the likelihood of clinically significant prostate cancer (csPCa). We explain the zone-dependent rules, sequence-specific criteria, and key improvements of v2.1, including its refined transition zone assessment. Furthermore, we contextualize its clinical impact with current evidence and discuss emerging trends, such as biparametric MRI and artificial intelligence, that are shaping the future of prostate imaging.

ABSTRACT
Stillbirth is more common in low and middle-income countries like Tanzania. The risk of stillbirth is high in intrauterine growth-restricted fetuses. Placental insufficiency is associated with intrauterine growth restriction as well as the risk of both preeclampsia and stillbirth. We present a 36-year-old preeclamptic woman with sonographic features of placental insufficiency, intrauterine growth restriction, whose obstetric outcome was a stillbirth. The obstetric Doppler findings indicated severe placental insufficiency and absent end-diastolic flow in the middle cerebral artery, which we consider a sign of impending fetal death/stillbirth.

ABSTRACT
Lissencephaly is a rare and severe neurodevelopmental disorder caused by defective neuronal migration during the first trimester of embryogenesis, resulting in a smooth cerebral cortex lacking normal gyri and sulci. We present case report of a female neonate, less than one month old who was admitted to NICU at the Komfo Anokye Teaching Hospital in Kumasi, Ghana, presenting with neonatal jaundice and an elongated cranial shape. Further investigations, with a Magnetic Resonance Imaging (MRI) on a Toshiba Achieva 1.5 Tesla scanner, revealed bilateral frontoparietal and temporal lissencephaly, absence of the septum pellucidum, basal ganglia fusion with poor internal capsule definition, and ventricular dilation. These findings were consistent with lissencephaly, likely resulting from tubulinopathy, coexisting with lobar holoprosencephaly. A multidisciplinary approach was employed in the management of this neonate. This case highlights the importance of MRI in diagnosing and understanding complex congenital brain anomalies, particularly in resource-limited settings like Ghana.