Rangpur Medical College

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by impairments in social interaction, communication, and behavior. Accurately recognizing the behaviors of children with ASD is crucial for diagnosis and early intervention. Traditional clinician- or caregiver-based approaches are often time-consuming, subjective, and delayed. In contrast, computer vision, along with modern deep learning (DL)-based methods, offers promising alternative solutions. However, a significant obstacle for this purpose is the lack of large-scale, publicly available datasets of ASD-related behaviors. To address the gap, we present the Autism Action Dataset (AAD), a novel publicly available video dataset including 202 manually annotated videos with seven distinct ASD-related action classes: Jumping, Arm Flapping, Joint Attention, Spinning, Head Banging, Head Shaking, and Imitation. These videos were collected from diverse YouTube sources to ensure variability in environment and subject condition.We evaluated the dataset using several state-of-the-art convolutional neural network (CNN)-based action recognition benchmarks, including CNN-LSTM, Inflated 3D (I3D), SlowFast, and Expandable 3D (X3D), as well as transformer-based attention benchmarks such as Video Vision Transformer (ViViT) and Multiscale Vision Transformer (MViT). Additionally, we evaluated pre-trained models fine-tuned on our dataset, as well as models trained from scratch under 5-fold stratified cross-validation. Among fine-tuned pre-trained methods, X3D achieved the best performance, reaching an accuracy of 84.18% along with a 79.82% F1-score. When trained from scratch, I3D achieved the highest performance with an accuracy of 64.29% with a 59.29% F1-score. These results demonstrate both the value of the AAD dataset as a publicly available benchmark and the efficacy of the I3D and X3D methods for ASD behavior analysis. They provide a solid foundation for future research and the development of assistive diagnostic tools. The dataset, metadata, and corresponding benchmark result will be publicly available at https://github.com/reyadhasan605/Autism-Action-Dataset.

Globally, undergraduate students exhibit a higher tendency towards tobacco and other substance use. Although there are almost 50,000 undergraduate medical students studying in Bangladesh currently, evidence regarding the prevalence of tobacco other substance use among them is very limited, let alone the predictors. This research aims to investigate the prevalence of tobacco or other substance use among medical students in Bangladesh and to investigate the factors associated with tobacco or other substance use. This cross-sectional study was conducted between May 2023 and August 2024 in sixteen medical colleges in Bangladesh. A total of 1,129 medical students were selected by convenience sampling, and data were collected using an online self-reported semi-structured questionnaire. Descriptive statistics, chi-square test, t-test, and multiple logistic regression model were used to report the study findings. The mean age of the students was 22.3 (± 2) years, with slightly higher female preponderance (52.3

Background: Hypothyroidism is the most common thyroid problem in Bangladesh and worldwide. Autoimmune thyroid disease is the most common cause of hypothyroidism. Levothyroxine (LT4) dose requirements vary among patients and may be influenced by thyroid autoantibody status and metabolic comorbidities such as type 2 diabetes mellitus (T2DM) to achieve euthyroidism. Objective: To assess the relationship between thyroid autoantibody status and levothyroxine dose requirements in hypothyroid patients with or without T2DM. Methods: This cross-sectional study was conducted at the Department of Biochemistry, BIRDEM General Hospital, from July 2024 to June 2025 on 100 patients with hypothyroidism recruited from outpatient department of endocrinology, BIRDEM, Dhaka, and divided into two groups: Group 1 (Hypothyroid patients with T2DM, n=50) and Group 2 (Hypothyroid patients without T2DM, n=50). Data on demographic, physical, biochemical variables and dose of levothyroxine were collected. Thyroid peroxidase (TPO) antibody, thyroglobulin (Tg) antibody, TSH and HbA1c were measured. Statistical analysis was performed using SPSS version 29.0. Results: Among 100 patients with hypothyroidism, mean age was 49.66±11.19 years in Group 1 and 45.84±14.4 years in Group 2. TPO antibody positivity was significantly lower in hypothyroid patients with T2DM (36%) compared to those without T2DM (66%) (p=0.003).The mean levothyroxine dose was substantially higher among anti-TPO positive patients than anti-TPO negative patients (67.6±28.0 μg vs. 40.6±17.8 μg, p<0.001), with a similar pattern observed for anti-Tg positivity (74.1±32.3 μg vs. 46.7±20.3 μg, p<0.001). Conclusion: Thyroid autoantibody positivity is strongly associated with higher levothyroxine dose requirements in hypothyroid patients, irrespective of diabetes status. Recognition of thyroid autoimmunity is essential for individualized dose optimization. J Rang Med Col.2026 Mar;11(1):177-182

The incidence of oral cancer is significantly increased by a lack of knowledge about its risk factors and early symptoms. This study aimed to assess patients’ knowledge and awareness of oral cancer and precancerous lesions attending the outpatient department of Dhaka Dental College Hospital. A cross-sectional study was conducted among 342 patients aged 18–76 years using convenience sampling. A semi-structured questionnaire assessed knowledge of risk factors, signs and symptoms, treatment, awareness of screening, and sources of information. Continuous data were dichotomized into ‘good’ and ‘poor’ knowledge levels using median splits—an approach that may limit statistical interpretability. A score of ≥ 50

Background Renal stone disease constitutes a significant global urological burden, with percutaneous nephrolithotomy (PCNL) serving as the gold-standard treatment for large calculi. Mini-PCNL has emerged to reduce morbidity while preserving efficacy; however, comparative prospective evidence from South Asian settings, including Bangladesh, remains scarce. This study aimed to compare the clinical outcomes, including efficacy and morbidity, of mini-PCNL and standard PCNL in the management of renal stones. Methods A prospective comparative study was conducted at the Department of Urology, Chittagong Medical College Hospital, Chattogram, Bangladesh, and two private hospitals in the same region from May 2019 to April 2020. A total of 100 adult patients with renal calculi (1-3 cm) were recruited through nonprobability consecutive sampling and allocated equally into two groups on the basis of surgeon discretion and intraoperative judgment: standard PCNL (24-28 Fr, n = 50) and mini-PCNL (14-20 Fr, n = 50). Operative parameters, postoperative pain (visual analog scale (VAS) score), hemoglobin drop, hospital stay, stone-free rate, renal function, and complications were compared between the groups. Data were analyzed using independent t-tests and chi-square tests, with p < 0.05 considered statistically significant. Results Baseline characteristics, including age, sex, and stone size, were comparable between the groups. Mini-PCNL demonstrated a significantly shorter access time (4.28 ± 0.57 vs 7.87 ± 0.39 min, p < 0.001) and fluoroscopy time (3.75 ± 0.51 vs 5.83 ± 0.65 min, p < 0.001) but a longer operative time (94.53 ± 13.56 vs 78.60 ± 17.80 min, p < 0.001). Postoperative pain (VAS score: 5.52 ± 1.24 vs 7.04 ± 1.09, p < 0.001), hemoglobin drop (1.02 ± 0.32 vs 1.31 ± 0.79 g/dL, p = 0.018), and length of hospital stay (2.32 ± 0.46 vs 3.13 ± 0.72 days, p < 0.001) were significantly lower in the mini-PCNL group. Stone-free rates, renal function outcomes, and complication rates were comparable between the two techniques. Conclusions In this study, mini-PCNL was associated with lower perioperative morbidity, shorter hospital stay, and reduced fluoroscopy exposure compared with standard PCNL, while stone clearance was comparable between the two groups. However, operative time was longer with mini-PCNL. Given the nonrandomized allocation and potential for selection bias, these findings should be interpreted with caution. Larger randomized multicenter studies are needed to confirm these observations.