Search

Convolutional Neural Networks (CNNs) are a class of deep neural networks that have proven highly effective in areas such as image and video recognition. CNNs typically include several types of layers, such as convolutional layers, activation layers, pooling layers, and fully connected layers, all of which contribute to the network's ability to recognize patterns and features. The pooling layer, which often follows the convolutional layer, is crucial for reducing computational complexity by performing down-sampling while maintaining essential features. This layer's role in balancing the symmetry of information across the network is vital for optimal performance. However, the choice of pooling method is often based on intuition, which can lead to less accurate or efficient results. This research compares various standard pooling methods (MAX and AVERAGE pooling) on standard datasets (MNIST, CIFAR-10, and CIFAR-100) to determine the most effective approach in preserving detail, performance, and overall computational efficiency while maintaining the symmetry necessary for robust CNN performance. [ABSTRACT FROM AUTHOR]

Background: Chromoblastomycosis (CBM) is a chronic infection of skin and subcutaneous tissue. CBM cases have been reported in local literature from Pakistan with heterogenous demographic, diagnostic and therapeutic information. The objective of this study is to share the experience of CBM from a large tertiary care hospital laboratory in Pakistan. Method: This was a retrospective observational study. Histopathology and microbiology data of suspected CBM between 2016 and 2022 was retrieved. Patients' demographics, site of involvement, histopathological findings and positive microbiology cultures were assessed. Literature search on Google Scholar, PubMed and PakMediNet was done between 1990 and 2023 with multiple terms. Result: A total of 16 CBM cases were identified; 14 were histopathology positive and two were both histopathology and culture positive. The median age was 21 years, and 11 patients were male. The predominant site was lower extremities followed by the face. Severe acanthosis, hyperkeratosis and granuloma with sclerotic bodies were observed in all histopathology slides. Alternaria spp. and Phialophora spp. were isolated from two culture‐positive cases. A total of nine cases of CBM were reported from Pakistan in PubMed non‐indexed journal. Conclusion: CBM is not a commonly thought of disease when evaluating skin lesions in Pakistan. A high index of suspicion when assessing patients who may have a history of trauma, exposure to soil and suggestive lesions is reasonable. An integrated approach between clinicians, histopathologist and microbiologist is required to do early identification and therapeutic interventions. [ABSTRACT FROM AUTHOR]

Salmonella enterica serotype Typhi is a major pathogen affecting populations from low- and middleincome countries that generally lack clean, potable water and good sanitary disposal systems (1). The global incidence [...]

Medical imaging is essential in contemporary healthcare as it assists in the identification of diseases, development of treatment strategies, and ongoing monitoring of patients. Over the years, deep learning (DL)techniques have emerged as a transformative force in medical image reconstruction, enabling the generation of high-quality images from noisy, incomplete, or under-sampled data. This review paper provides a comprehensive survey of recent advancements and applications of deep learning methods in medical image reconstruction. The key challenges in medical image reconstruction include issues related to reconstruction accuracy, noise sensitivity, and data limitation. A variety of deep learning models and their combinations are suitable for medical image reconstruction due to their unique capabilities, such as spatial hierarchy capture, adversarial learning, and other features, which allow them to address the complexities and challenges associated with medical image reconstruction. The paper analyses the key contributions of DL-based approaches in different imaging modalities, including computed tomography (CT) and magnetic resonance imaging (MRI). DL techniques are enabling image reconstruction in specialized medical fields like neuroimaging and cardiac imaging, but practitioners face challenges in training complex models and understanding their results. Finally, future research directions are suggested to improve the key limitations highlighted in this survey study. [ABSTRACT FROM AUTHOR]

Background: Correct identification of clinically significant mold is becoming critical and cannot be relied only on phenotypic methods; hence, there is a dire need to develop an algorithm of workflows and capacity for molecular identification. This study shares the experience of DNA sequencing for invasive mold identification isolated and compared with phenotypic identification. Methods: This study was conducted at the microbiology laboratory, Aga Khan University, Karachi, Pakistan, and the Mycotic Disease Branch, Centers for Disease Control and Prevention, USA. Filamentous molds isolated from clinical specimens during January 2012-April 2013 were initially identified through phenotypic characteristics. Pan-fungal polymerase chain reaction targeting the internal transcribed spacer region of the ribosomal cistron and the D1/D2 domains of the 28S ribosomal cistron was performed. Sequencer version was utilized to edit and align the DNA sequences, and then sequences were identified using BLAST. The correlation between phenotypic and molecular identification was evaluated. Results: Gene sequencing identified 50% of clinical isolates as one of the Aspergillus species, followed by Mucorales 29%, Fusarium species 17%, and Cladosporium spp. 4%. Overall, 50% of clinical isolates were identified correctly till the species level by conventional methods. Phenotypic correlation with genotype till genus was 42%, while two isolates were wrongly identified phenotypically. Excellent agreement (100%) till species level between phenotypic and molecular identification for Aspergillus, while Mucorales had an agreement of 71%. Conclusion: Traditional phenotypic identification methods for filamentous molds had a good correlation with sequencing up to genus level identification; however, were not reliable up to species level. [ABSTRACT FROM AUTHOR]

Objective: To examine internal organ tissues and placentas of stillbirths for various pathogens. Design: Prospective, observational study. Settings: Three study hospitals in India and a large maternity hospital in Pakistan. Population: Stillborn infants delivered in a study hospital. Methods: A prospective observational study. Main outcome measures: Organisms identified by pathogen polymerase chain reaction (PCR) in internal organs and placental tissues of stillbirths. Results: Of 2437 stillbirth internal tissues, 8.3% (95% CI 7.2–9.4) were positive. Organisms were most commonly detected in brain (12.3%), cerebrospinal fluid (CSF) (9.5%) and whole blood (8.4%). Ureaplasma urealyticum/parvum was the organism most frequently detected in at least one internal organ (6.4% of stillbirths and 2% of all tissues). Escherichia coli/Shigella was the next most common (4.1% one or more internal organ tissue sample and 1.3% of tissue samples), followed by Staphylococcus aureus in at least one internal organ tissue (1.9% and 0.9% of all tissues). None of the other organisms was found in more than 1.4% of the tissue samples in stillbirths or more than 0.6% of the internal tissues examined. In the placenta tissue, membrane or cord blood combined, 42.8% (95% CI 40.2–45.3) had at least one organism identified, with U. urealyticum/parvum representing the most commonly identified (27.8%). Conclusions: In about 8% of stillbirths, there was evidence of a pathogen in an internal organ. Ureaplasma urealyticum/parvum was the most common organism found in the placenta and in the internal tissues, especially in the fetal brain. [ABSTRACT FROM AUTHOR]

INTRODUCTION: In research projects such as vaccine trials, accurate and complete surveillance of all outcomes of interest is critical. In less developed countries where the private sector is the major health-care provider, the private sector must be included in surveillance systems in order to capture all disease of interest. This, however, poses enormous challenges in practice. The process and outcome of recruiting private practice clinics for surveillance in a vaccine trial are described. METHODS: The project started in January 2002 in two urban squatter settlements of Karachi, Pakistan. At the suggestion of private practitioners, a phlebotomy team was formed to provide support for disease surveillance. Children who had a reported history of fever for more than three days were enrolled for a diagnosis. RESULTS: Between May 2003 and April 2004, 5540 children younger than 16 years with fever for three days or more were enrolled in the study. Of the children, 1312 (24%) were seen first by private practitioners; the remainder presented directly to study centres. In total, 5329 blood samples were obtained for microbiology. The annual incidence of Salmonella typhi diagnosed by blood culture was 407 (95% confidence interval (95% CI), 368-448) per 100 000/year and for Salmonella paratyphi A was 198 (95% CI, 171-227) per 100 000/year. Without the contribution of private practitioners, the rates would have been 240 per 100 000/year (95% CI, 211-271) for S. typhi and 114 (95% CI, 94-136) per 100 000/year for S. paratyphi A. CONCLUSION: The private sector plays a major health-care role in Pakistan. Our experience from a surveillance and burden estimation study in Pakistan indicates that this objective is possible to achieve but requires considerable effort and confidence building. Nonetheless, it is essential to include private health care providers when attempting to accurately estimate the burden of disease in such settings.

To the Editor: Naegleria fowleri. a free-living ameba, causes acute, fulminant, fatal primary amebic meningoencephalitis (PAM) in persons with history of recreational activities in warm freshwater (1,2). During 2008-2009, thirteen [...]

Background We identified pathogens found in internal organs and placentas of deceased preterm infants cared for in hospitals in India and Pakistan. Methods Prospective, observational study conducted in delivery units and neonatal intensive care units. Tissue samples from deceased neonates obtained by minimally invasive tissue sampling and placentas were examined for 73 different pathogens using multiplex polymerase chain reaction (PCR). Results Tissue for pathogen PCR was obtained from liver, lung, brain, blood, cerebrospinal fluid, and placentas from 377 deceased preterm infants. Between 17.6% and 34.1% of each type of tissue had at least 1 organism identified. Organism detection was highest in blood (34.1%), followed by lung (31.1%), liver (23.3%), cerebrospinal fluid (22.3%), and brain (17.6%). A total of 49.7% of the deceased infants had at least 1 organism. Acinetobacter baumannii was in 28.4% of the neonates compared with 14.6% for Klebsiella pneumoniae , 11.9% for Escherichia coli/Shigella , and 11.1% for Haemophilus influenzae. Group B streptococcus was identified in only 1.3% of the neonatal deaths. A. baumannii was rarely found in the placenta and was found more commonly in the internal organs of neonates who died later in the neonatal period. The most common organism found in placentas was Ureaplasma urealyticum in 34% of the samples, with no other organism found in >4% of samples. Conclusions In organ samples from deceased infants in India and Pakistan, evaluated with multiplex pathogen PCR, A. baumannii was the most commonly identified organism. Group B streptococcus was rarely found. A. baumannii was rarely found in the placentas of these deceased neonates. [ABSTRACT FROM AUTHOR]

Primary amebic meningoencephalitis (PAM) is a fatal disease caused by the thermotolerant free-living ameba Naegleria fowleri. Found worldwide in moist soil and freshwater, these amebae proliferate during summer when ambient [...]