Mahidol University's Institutional Repository
คลังสารสนเทศสถาบันของมหาวิทยาลัยมหิดล
"Wisdom Repository You Discover"
To collect Mahidol University's academic publications and intellectual properties more than 39 faculties
To present over 50,000 items of information in digital formats
To make it easy to access to all information at anytime, anywhere
Communities in Mahidol IR
Select a community to browse its collections.
Recent Submissions
Human Fall Detection Through Velocity-Driven Temporal Deep Learning
(2025-01-01) Ahmed N.; Hasan M.R.; Shakib F.H.; Rabbani H.; Rabbi R.; Khan S.T.; Zereen A.N.; Ahmed N.; Mahidol University
Accidental falls remain one of the most frequent and dangerous incidents among older adults, often leading to hospitalization, long-term disability, and loss of independence. This research presents a novel, non-wearable, vision-based system that addresses the challenge of real-time fall detection and instability recognition in home environments. The system uses monocular-camera-captured video to track human motion and classify it into three categories: normal, fall, or unstable. By employing YOLOv8 for keypoint extraction and a two-stage deep learning framework, the model predicts human movement velocity and classifies activities based on temporal skeleton keypoint dynamics. The two-stage framework consists of two separate LSTM-based models: the first stage predicts velocity from keypoint sequences using regression, and the second stage classifies motion into three categories. Evaluation on a primary dataset of 3449 annotated high-resolution video clips demonstrates a 90% classification accuracy. Compared to baseline methods, the proposed system improves the accuracy of fall detection, showing strong potential for practical deployment in elderly care settings. The approach offers a scalable and unobtrusive solution for continuous monitoring, eliminating the need for wearable devices.
NPC: Automated Tool for Detecting and Explaining ChatGPT-Generated Programs
(2026-01-01) Saeheng P.; Boongaree N.; Sriwilailak C.; Ragkhitwetsagul C.; Racharak T.; Chuangsuwanich E.; Saeheng P.; Mahidol University
The adoption of Large Language Models (LLMs) is rising in programming education, which raises concerns regarding academic dishonesty and a lack of trustworthiness in students’ programming submissions. There are recent automated techniques and tools for classifying submitted code as generated by LLMs or created by students. However, they lack an explanation of their decision, which educators often require to make informed decisions. This paper presents NPC, an approach for detecting and explaining code snippets generated by ChatGPT, employing machine learning and our proposed local neighborhood sampling strategy to build post-hoc explainability in artificial intelligence (AI). We develop our approach as a web application that not only detects ChatGPT-generated code but also provides educators with explanations in an easy-tounderstand format for each classification. The evaluation found that the explanations were clear and helpful, reinforcing the tool’s potential to support academic integrity in programming education. The video demonstration of the tool is available at https://bit.ly/ase25-npc-demo. The tool’s source code is publicly available at https://github.com/pachanitha/NPC Project.
Associations Between Caregiver Smartphone Use in a Child's Presence and Motor Skills and Executive Function in Preschoolers: SUNRISE International Study
(2026-07-01) Toledo-Vargas M.; Chong K.H.; Engberg E.; Jáuregui A.; Martins C.; Byambaa A.; López-Gil J.F.; Abdeta C.; Mwase-Vuma T.; El Hamdouchi A.; Lubree H.; Aoko O.A.; Chelly M.S.; Chia M.Y.H.; Ghofranipour F.; Gonzalez-Santamaria J.; Jarani J.; Katewongsa P.; Kontsevaya A.; Singh P.; Subedi N.; Teo W.P.; Tiongco M.M.; Trampa K.; Turab A.; Užičanin E.; Veldman S.L.C.; Okely A.D.; Toledo-Vargas M.; Mahidol University
Background: The preschool years are a crucial period for development. Stable environments and responsive caregivers support children's cognitive and motor development, two interrelated and essential domains. Caregiver smartphone use in front of children may reduce attention and responsiveness, which have been negatively associated with young children's health and development. We examined associations between the frequency of caregiver smartphone use in a child's presence and motor skills and executive functions, and whether these associations vary by country income level. Methods: We analysed cross-sectional data from 27 countries participating in the SUNRISE International Study. Caregivers reported the frequency of smartphone use in the child's presence across five scenarios: mealtime, playtime, travel, walk and bedtime routines. Children's motor skills were assessed using four established tests. Early Years Toolbox games were used to assess children's visual–spatial working memory and inhibition. Mixed-effects linear regression models were used to assess the associations, including interaction terms to test variation by country income level. Models were adjusted for the child's sex, age, daily screen time, time spent outdoors, sleep duration and the caregiver's highest level of education. Results: The analytical sample included 2232 preschoolers (mean age = 4.2 ± 0.6 years, 50.9% girls). In fully adjusted models, the frequency of caregiver smartphone use in a child's presence was not associated with gross motor skills, fine motor skills, visual–spatial working memory or inhibition (all p > 0.05). Results did not differ by country income level. Conclusions: Findings suggest that the frequency of caregiver smartphone use in a child's presence alone may not be associated with performance on motor skills and executive function. There is a need for more sensitive measures that capture the frequency, duration and context of interruptions and more longitudinal studies that examine motor development and cognition. Future research should also account for socioeconomic and demographic diversity, environmental factors and cultural context when assessing such associations.
Carbapenemases: epidemiology, detection and management in a changing global landscape
(2026-06-01) Macesic N.; Harris P.N.A.; Mo Y.; Gomez-Simmonds A.; Macesic N.; Mahidol University
Carbapenemase-producing Gram-negative bacteria are a growing threat to last-line beta-lactam (BL) therapy and are now established across Enterobacterales, Pseudomonas aeruginosa and Acinetobacter baumannii. Since the first reports in the 1980s, carbapenemase genes have disseminated internationally through plasmids, transposons and integrons and have become embedded in high-risk clones, with contemporary epidemiology dominated by Klebsiella pneumoniae carbapenemases (KPC), New Delhi metallo-beta-lactamase, Verona integron-encoded, imipenemase and oxacillinase (OXA)-type enzymes. The past decade has been marked by spread beyond hospitals, rising metallo-beta-lactamase (MBL) prevalence, and convergence of resistance with hypervirulence in some lineages. This Review summarizes carbapenemase classification, genetic contexts and epidemic clones. It describes regional distribution patterns alongside One Health drivers linking healthcare, community, animal and environmental reservoirs. We outline a pragmatic diagnostic framework spanning screening, confirmatory phenotypic assays, rapid lateral flow and molecular platforms, and whole genome sequencing for surveillance and outbreak investigation, emphasizing the clinical value of early mechanism identification for both infection control and targeted therapy. Treatment is reviewed in a mechanism-directed manner: newer BL/beta-lactamase inhibitor combinations are central for serine carbapenemases (including KPC and many OXA-48-like producers), whereas MBL producers require alternative strategies such as aztreonam-based combinations or cefiderocol. Options remain limited for carbapenemase-producing P. aeruginosa and A. baumannii, although sulbactam-durlobactam and pipeline agents are expanding the therapeutic landscape. We highlight the widening gap between disease burden and access to rapid diagnostics and novel therapies, particularly in high-burden low- and middle-income settings. Finally, we outline the bundled infection prevention and antimicrobial stewardship interventions needed to contain transmission and preserve the effectiveness of novel agents.
Improved Prediction of Cardiovascular Events Using Serial Cardio-Ankle Vascular Index (CAVI) Measurements: A 10-Year Prospective Cohort Study
(2026-06-01) Limpijankit T.; Vathesatogkit P.; Matchariyakul D.; Thongmung N.; Siriyotha S.; Thakkinstian A.; Sritara P.; Limpijankit T.; Mahidol University
Background: Most studies evaluating the cardio-ankle vascular index (CAVI) as a marker of arterial stiffness are cross-sectional, limiting insights into long-term vascular changes. We investigated whether serial CAVI measurements improve the prediction of cardiovascular (CV) events beyond a single baseline value. Methods: The Electricity Generating Authority of Thailand (EGAT) study is a prospective cohort with 5-year follow-up intervals. Participants with prior coronary artery disease (CAD) or stroke were excluded. Demographic, clinical, laboratory data, medication use, and CAVI were collected. Serial CAVI was analyzed as a time-varying covariate in Cox proportional hazards models, incorporating baseline, 5-year, and 10-year measurements. The primary composite CV outcomes comprised CAD, stroke, or CV death. Cox proportional hazards models assessed associations between baseline or serial CAVI and CV outcomes, adjusted for conventional risk factors. Results: Among 3913 participants (mean age 49.0 ± 10.8 years; 73.4% male; BMI 24.3 ± 3.6 kg/m2), mean CAVI increased from 7.7 ± 1.1 to 8.2 ± 1.3 over 10 years (p < 0.001). During a median follow-up of 9.1 ± 3.1 years, 0.8% experienced composite CV events. Serial CAVI was independently associated with CV outcomes (HR 1.47; 95% CI 1.06–2.03; p = 0.019), whereas baseline CAVI was not (HR 1.14; 95% CI 0.93–1.40; p = 0.22). Model indices (ΔAIC = 37.1; ΔBIC = 32.7) supported superior predictive performance for serial CAVI. Conclusion: Serial CAVI measurements better predict long-term CV events than a single baseline value. Longitudinal CAVI monitoring may enhance CV risk stratification and support preventive cardiovascular care.