Assuntos

RESUMO

BACKGROUND: Timely diagnosis plays a critical role in determining melanoma prognosis, prompting the development of deep learning models to aid clinicians. Questions persist regarding the efficacy of clinical images alone or in conjunction with dermoscopy images for model training. This study aims to compare the classification performance for melanoma of three types of CNN models: those trained on clinical images, dermoscopy images, and a combination of paired clinical and dermoscopy images from the same lesion. MATERIALS AND METHODS: We divided 914 image pairs into training, validation, and test sets. Models were built using pre-trained Inception-ResNetV2 convolutional layers for feature extraction, followed by binary classification. Training comprised 20 models per CNN type using sets of random hyperparameters. Best models were chosen based on validation AUC-ROC. RESULTS: Significant AUC-ROC differences were found between clinical versus dermoscopy models (0.661 vs. 0.869, p < 0.001) and clinical versus clinical + dermoscopy models (0.661 vs. 0.822, p = 0.001). Significant sensitivity differences were found between clinical and dermoscopy models (0.513 vs. 0.799, p = 0.01), dermoscopy versus clinical + dermoscopy models (0.799 vs. 1.000, p = 0.02), and clinical versus clinical + dermoscopy models (0.513 vs. 1.000, p < 0.001). Significant specificity differences were found between dermoscopy versus clinical + dermoscopy models (0.800 vs. 0.288, p < 0.001) and clinical versus clinical + dermoscopy models (0.650 vs. 0.288, p < 0.001). CONCLUSION: CNN models trained on dermoscopy images outperformed those relying solely on clinical images under our study conditions. The potential advantages of incorporating paired clinical and dermoscopy images for CNN-based melanoma classification appear less clear based on our findings.

Assuntos

RESUMO

Me gustaría comenzar este editorial con una perspectiva positiva; sin embargo, la gravedad de la situación requiere abordar los desafíos de manera clara con el objetivo de fomentar respuestas efectivas. Las enfermedades crónicas no transmisibles son una amenaza real. En este momento, su control debería ser la máxima prioridad de la salud pública en regiones como Latinoamérica. No obstante, al acercarnos al cuarto de siglo del siglo XXI, observamos que nuestro progreso como sociedad en el control de las enfermedades crónicas no transmisibles aún es insuficiente.

I would like to begin this editorial with a positive outlook; however, the gravity of the situation requires addressing the challenges clearly in order to foster effective responses. Chronic noncommunicable diseases are a real threat. At this point in time, their control should be the top public health priority in regions such as Latin America. However, as we approach the quarter century of the 21st century, we observe that our progress as a society in the control of chronic noncommunicable diseases is still insufficient.

Assuntos

Assuntos

RESUMO

Visualizing anatomical structures and functional processes in three dimensions (3D) are important skills for medical students. However, contemplating 3D structures mentally and interpreting biomedical images can be challenging. This study examines the impact of a new pedagogical approach to teaching neuroanatomy, specifically how building a 3D-model from oil-based modeling clay affects learners' understanding of periventricular structures of the brain among undergraduate medical students in Colombia. Students were provided with an instructional video before building the models of the structures, and thereafter took a computer-based quiz. They then brought their clay models to class where they answered questions about the structures via interactive response cards. Their knowledge of periventricular structures was assessed with a paper-based quiz. Afterward, a focus group was conducted and a survey was distributed to understand students' perceptions of the activity, as well as the impact of the intervention on their understanding of anatomical structures in 3D. Quiz scores of students that constructed the models were significantly higher than those taught the material in a more traditional manner (P < 0.05). Moreover, the modeling activity reduced time spent studying the topic and increased understanding of spatial relationships between structures in the brain. The results demonstrated a significant difference between genders in their self-perception of their ability to contemplate and rotate structures mentally (P < 0.05). The study demonstrated that the construction of 3D clay models in combination with autonomous learning activities was a valuable and efficient learning tool in the anatomy course, and that additional models could be designed to promote deeper learning of other neuroanatomy topics. Anat Sci Educ 11: 137-145. © 2017 American Association of Anatomists.

Assuntos

RESUMO

Introducción: Las tecnologías de información y la facilidad de acceso a grandes volúmenes de datos "big data" están revolucionando la epidemiología. Objetivo: Analizar datos del sistema integral de Información "SISPRO" del Ministerio de Salud para obtener los motivos más frecuentes de consulta en otorrinolaringología en consulta externa y urgencias. Diseño: Estudio descriptivo ecológico. Metodología: En SISPRO se usaron los códigos CIE-10 para analizar los datos de atención en salud para patologías de oído (H60-H959) y vía aérea superior (J00-J399). Resultados: Los principales motivos de consulta incluyen: infecciones de la vía aérea superior (rinofaringitis, faringitis, amigdalitis, otitis media aguda e infecciones no especificadas) con 14.935.729 atenciones, 7.877.189 personas atendidas y un promedio de 1,9 consultas por persona; Rinitis (1.760.055 atenciones, 899.349 personas y 1,96 consultas/persona); Vértigo (1.632.012 atenciones, 686.470 personas y 2,38 consultas/persona); Cerumen impactado (554.744 atenciones, 296.324 personas y 18,76 consultas/persona); Hipoacusia (853.047 atenciones, 366.037 personas y 2,33 consultas/persona). La relación entre el número de atenciones en consulta externa vs urgencias es de 281,66 (hipoacusia), 126,23 (Rinitis), 37,2 (Cerumen impactado), 10,73 (vértigo) y 6,9 (infecciones). Conclusión: La epidemiología digital es una herramienta útil en la práctica de ORL. Las enfermedades infecciosas son la principal causa de atención en ORL.

Introduction: Information technologies and the ease of access to large volumes of data "big data" are revolutionizing epidemiology. Objective: Analyze data from the Comprehensive Information System "SISPRO" of the Ministry of Health, to obtain the most frequent reasons for consultation in otorhinolaryngology in outpatient and emergency departments. Design: ecological descriptive study. Methods: In SISPRO, ICD-10 codes were used to analyze health care data for ear (H60-H959) and upper airway (J00-J399) pathologies. Results: The main reasons for consultation included: upper airway infections (rhinopharyngitis, pharyngitis, tonsillitis, acute otitis media and unspecified infections) with 14,935,729 attentions, 7,877,189 people attended and an average of 1.9 consultations per person; Rhinitis (1,760,055 attentions, 899,349 people and 1.96 consultations / person); Vertigo (1,632,012 attentions, 686,470 people and 2,38 consultations / person; Cerumen impacted (554,744 attentions, 296,324 people and 18,76 consultations / person); Hearing loss (853,047 attentions, 366,037 people and 2,33 consultations / person). The ratio between the number of visits in the outpatient clinic vs. emergencies is 281.66 (hearing loss), 126.23 (rhinitis), 37.2 (impacted earwax), 10.73 (vertigo) and 6.9 (infections). Conclusion: Digital epidemiology is an useful tool in ENT. Infectious diseases are the main cause of consultation in ENT.

Assuntos