Bibliographic citations
This is an automatically generated citacion. Modify it if you see fit
Alva, A., (2022). Soft-Warmi: evaluación del software automático de diagnóstico para vaginosis bacteriana (VB) [Universidad Peruana Cayetano Heredia]. https://hdl.handle.net/20.500.12866/12938
Alva, A., Soft-Warmi: evaluación del software automático de diagnóstico para vaginosis bacteriana (VB) []. PE: Universidad Peruana Cayetano Heredia; 2022. https://hdl.handle.net/20.500.12866/12938
@mastersthesis{renati/910512,
title = "Soft-Warmi: evaluación del software automático de diagnóstico para vaginosis bacteriana (VB)",
author = "Alva Mantari, Alicia Katherine",
publisher = "Universidad Peruana Cayetano Heredia",
year = "2022"
}
Title: Soft-Warmi: evaluación del software automático de diagnóstico para vaginosis bacteriana (VB)
Authors(s): Alva Mantari, Alicia Katherine
OCDE field: http://purl.org/pe-repo/ocde/ford#1.02.00; http://purl.org/pe-repo/ocde/ford#3.02.02; http://purl.org/pe-repo/ocde/ford#3.03.08
Issue Date: 2022
Institution: Universidad Peruana Cayetano Heredia
Abstract: La vaginosis bacteriana (VB) puede considerarse un problema de salud pública debido a la cantidad de mujeres afectadas por ella. Esto es debido a que es asintomática, en la mayoría de casos. Los problemas asociados a ella corresponden principalmente a la vulnerabilidad a la salud femenina al contraerla. La vaginosis bacteriana es alarmante principalmente durante los primeros tres meses de gestación, pues hace latente un riesgo de ruptura de membranas o abortos espontáneos entre otros. Cabe mencionar que las mujeres sexualmente activas además aumentan sus posibilidades de contraer y transmitir enfermedades de trasmisión sexual (ETS). Uno de los principales problemas asociados a VB, tiene que ver con el diagnostico mismo de la infección. El diagnóstico es altamente variable entre el personal entrenado para esta labor, razón por la cual es importante contar con herramientas que permitan un diagnóstico más rápido y confiable y menos variable referente a VB. Este estudio desarrollo una herramienta computacional para realizar tal diagnóstico, y además implemento un medio de comunicación masiva para hacer acceder remotamente a él. SoftWarmi es un software dedicado al diagnóstico de VB, de manera remota a partir de 10 fotos de la muestra. Este algoritmo se basa en el conteo de los morfotipos en la muestra de manera similar a lo realizado por el score de Nugent, el método utilizado de manera tradicional para el diagnóstico de VB. Se analizaron 59 láminas de hisopados vaginales. Las láminas estaban clasificadas por 30 positivas y 29 negativas a VB. De cada muestra han sido capturados 10 campos, seleccionadas por el Blga. Lorena Rojas, entrenada para tal fin anteriormente. De cada fotografía se han extraído un total de 500 objetos aprox. De cada objeto se han calculado un conjunto de 184 características cuantificables de forma, color, textura e iluminación. Posteriormente se diseñaron modelos estadísticos para el reconocimiento de Lactobacillus, Gardnerellas y Mobiluncus. Para Lactobacillus obtuvimos una sensibilidad de 97.63% y una especificidad de 99.10%. Para el reconocimiento de Gardnerella se alcanzó una sensibilidad de 93.64% y una especificidad de 97.40%. Por ultimo para Mobiluncus alcanzamos una sensibilidad de 96.23% y una especificidad de 95.51%. Luego se realizaron conteos de cada morfotipo de manera secuencial y excluyente, evaluándolas según la incidencia de existencia en la población general. Luego se calculó un modelo estadístico de diagnóstico por fotos, el cual alcanzo una sensibilidad de 100% y una especificidad de 96.15%, en el training set (50 fotos), y el en testing set en un conjunto de 66 fotos una sensibilidad de 94.44% y una especificidad de 100%. Al final se evaluaron 10 fotos por muestra, de 59 muestras estudiadas se alcanzó para este modelo una sensibilidad de 96.67% y una especificidad de 100%, con un porcentaje de agreement con los expertos de la universidad de Washington de 98.31% y un índice Kappa de 0.9661. Esperamos que este sea el inicio de una herramienta para lograr que la VB, sea analizada, diagnosticada y tratada en diferentes poblaciones que no poseen los medios para poder acceder sobre todo en las zonas más alejadas, pobres y remotas del país.
Bacterial vaginosis (BV) can be considered a public health problem due to the number of women affected by it. This is because it is asymptomatic, in most cases. The problems associated with it correspond mainly to the vulnerability of women's health to contracting it. Bacterial vaginosis is alarming mainly during the first three months of pregnancy, as it poses a latent risk of rupture of membranes or miscarriages, among others. It is worth mentioning that sexually active women also increase their chances of contracting and transmitting sexually transmitted diseases (STDs). One of the main problems associated with BV is the diagnosis of the infection itself. Diagnosis is highly variable among trained personnel, which is why it is important to have tools that allow a faster, more reliable and less variable diagnosis of BV. This study developed a computational tool to perform such diagnosis, and also implemented a means of mass communication to make it remotely accessible. SoftWarmi is a software dedicated to the diagnosis of BV, remotely from 10 photos of the sample. This algorithm is based on counting morphotypes in the sample in a similar way to Nugent's score, the traditional method used for BV diagnosis. Fifty-nine slides of vaginal swabs were analyzed. The slides were classified as 30 positive and 29 negatives for BV. From each sample, 10 fields were captured and selected by Blga. Lorena Rojas, trained for this purpose previously. A total of approximately 500 objects were extracted from each photograph. A set of 184 quantifiable features of shape, color, texture and illumination were calculated from each object. Subsequently, statistical models were designed for the recognition of Lactobacillus, Gardnerella and Mobiluncus. For Lactobacillus we obtained a sensitivity of 97.63% and a specificity of 99.10%. For the recognition of Gardnerella, we reached a sensitivity of 93.64% and a specificity of 97.40%. Finally, for Mobiluncus we achieved a sensitivity of 96.23% and a specificity of 95.51%. Counts of each morphotype were then carried out sequentially and in an exclusionary manner, evaluating them according to the incidence of occurrence in the general population. Then a statistical model of diagnosis by photos was calculated, which reached a sensitivity of 100% and a specificity of 96.15%, in the training set (50 photos), and in the testing set in a set of 66 photos a sensitivity of 94.44% and a specificity of 100%. In the end, 10 photos per sample were evaluated, and out of 59 samples studied, a sensitivity of 96.67% and a specificity of 100% were achieved for this model, with a percentage of agreement with the experts of the University of Washington of 98.31% and a Kappa index of 0.9661. We hope that this is the beginning of a tool to ensure that BV is analyzed, diagnosed and treated in different populations that do not have the means to access it, especially in the most remote, poor and remote areas of the country. (AU)
Bacterial vaginosis (BV) can be considered a public health problem due to the number of women affected by it. This is because it is asymptomatic, in most cases. The problems associated with it correspond mainly to the vulnerability of women's health to contracting it. Bacterial vaginosis is alarming mainly during the first three months of pregnancy, as it poses a latent risk of rupture of membranes or miscarriages, among others. It is worth mentioning that sexually active women also increase their chances of contracting and transmitting sexually transmitted diseases (STDs). One of the main problems associated with BV is the diagnosis of the infection itself. Diagnosis is highly variable among trained personnel, which is why it is important to have tools that allow a faster, more reliable and less variable diagnosis of BV. This study developed a computational tool to perform such diagnosis, and also implemented a means of mass communication to make it remotely accessible. SoftWarmi is a software dedicated to the diagnosis of BV, remotely from 10 photos of the sample. This algorithm is based on counting morphotypes in the sample in a similar way to Nugent's score, the traditional method used for BV diagnosis. Fifty-nine slides of vaginal swabs were analyzed. The slides were classified as 30 positive and 29 negatives for BV. From each sample, 10 fields were captured and selected by Blga. Lorena Rojas, trained for this purpose previously. A total of approximately 500 objects were extracted from each photograph. A set of 184 quantifiable features of shape, color, texture and illumination were calculated from each object. Subsequently, statistical models were designed for the recognition of Lactobacillus, Gardnerella and Mobiluncus. For Lactobacillus we obtained a sensitivity of 97.63% and a specificity of 99.10%. For the recognition of Gardnerella, we reached a sensitivity of 93.64% and a specificity of 97.40%. Finally, for Mobiluncus we achieved a sensitivity of 96.23% and a specificity of 95.51%. Counts of each morphotype were then carried out sequentially and in an exclusionary manner, evaluating them according to the incidence of occurrence in the general population. Then a statistical model of diagnosis by photos was calculated, which reached a sensitivity of 100% and a specificity of 96.15%, in the training set (50 photos), and in the testing set in a set of 66 photos a sensitivity of 94.44% and a specificity of 100%. In the end, 10 photos per sample were evaluated, and out of 59 samples studied, a sensitivity of 96.67% and a specificity of 100% were achieved for this model, with a percentage of agreement with the experts of the University of Washington of 98.31% and a Kappa index of 0.9661. We hope that this is the beginning of a tool to ensure that BV is analyzed, diagnosed and treated in different populations that do not have the means to access it, especially in the most remote, poor and remote areas of the country. (AU)
Link to repository: https://hdl.handle.net/20.500.12866/12938
Discipline: Informática Biomédica en Salud Global con mención en Informática en Salud
Grade or title grantor: Universidad Peruana Cayetano Heredia. Escuela de Posgrado Víctor Alzamora Castro
Grade or title: Maestro en Informática Biomédica en Salud Global con mención en Informática en Salud
Juror: Acho Mego, Segundo Cecilio; Cóndor Cámara, Daniel Flavio; Fonseca Arroyo, Pablo Alejandro
Register date: 20-Dec-2022
This item is licensed under a Creative Commons License