A metodologia ROC na avaliação de um modelo fuzzy de predição do estádio patológico do tumor de próstata / The ROC methodology in the evaluation of a fuzzy model to prediction of the prostate tumor's pathological stage

Nos últimos anos o aumento da incidência de casos de câncer de próstata configura-se como um importante problema de saúde pública e um desafio para a ciência médica. O objetivo deste trabalho é a avaliação do desempenho de um modelo matemático, desenvolvido por Silveira (2007) para predizer o estadiamento patológico do câncer de próstata, por meio da metodologia ROC (Receiver Operating Characteristic). O modelo consiste num sistema baseado em regras fuzzy (SBRF), que combina os dados pré-cirúrgicos – estado clínico, nível de PSA e grau de Gleason – acionando um conjunto de regras linguísticas, elaboradas com base nas informações presentes nos nomogramas já existentes. A saída do sistema fornece as possibilidades do indivíduo, com determinado quadro clínico, se enquadrar em cada um dos estádios de extensão do tumor: localizado, localmente avançado e metastático. Para a análise do poder discriminatório do modelo fuzzy como um teste de diagnóstico, foi construída, a partir das medidas de sensibilidade e especificidade, a curva ROC e calculada a área total sob a curva, como medida de desempenho. Além disso, foram obtidos (de duas maneiras distintas) os pontos de corte mais “adequados”, isto é, um limiar de decisão entre a doença estar totalmente localizada no interior da glândula prostática ou não. Dados reais de pacientes do Hospital de Clínicas da UNICAMP foram usados nos cálculos e a cirurgia– prostatectomia radical – foi adotada como padrão-ouro. Os resultados alcançados mostraram que o modelo fuzzy em questão pode vir a ser utilizado para discriminar câncer de próstata localizado.

In recent years, the increase in the incidence of prostate cancer has become a major public health problem and a challenge for medical science. The goal of this work is assessing the performance of a mathematical model, developed by Silveira (2007) to predict the pathological stage of the prostate cancer, through ROC methodology (Receiver Operating Characteristic). The model is a fuzzy rule based system, that combines pre-surgical data – clinical stage, PSA level and Gleason score – availing of a set of linguistic rules made with base on information of the existents nomograms. The output of the system provides the possibilities of the individual, with certain clinical features, be in each stage of the tumor extension: localized, advanced locally and metastatic. To analyze the discriminatory power of the fuzzy model as a diagnosis test, was constructed from the measures of sensitivity and specificity, the ROC curve and calculated the total area under the curve, as measure of performance. Moreover, were obtained (in two different ways) the cutoff points most “appropriate”, that is a threshold for deciding between the disease is fully localized within the prostate gland or not. Real data of patients from the Clinics Hospital of UNICAMP were used in the calculations and the surgery – radical prostatectomy – was used as gold standard. The results showed that the fuzzy model in question can be used to discriminate localized prostate cancer.

Environmental stress and temporomandibular disorder (TMD) among members of a public university in Brazil

Temporomandibular disorder (TMD) is one of several somatic manifestations of stress. Several studies have shown that some patients with TMD suffer more anxiety than individuals without this condition. The aim of this study was to identify individuals with TMD and correlate with environmental stress among members of a public university. A proportional stratified probabilistic procedure was used to select 455 subject’s age 17-63 yrs consisted of students, staff and professors. The subjects were enlisted after lunch. TMD were diagnosed using a screening questionnaire based on the American Academy of Orofacial Pain (AAOP). Presence of TMD was defined based on pain in chewing or speaking, tired maxillae and pain in the head, neck and/or teeth. Environmental stress was diagnosed by using the Social Readjustment Rating Scale (SRRS).Gender distribution was similar among undergraduate and graduate students, but there were significant differences in the sex ratio of university staff (66.7% were females) and professors (81.8% were males). Prevalence of TMD influenced by stress among women was significantly greater than in men of the same age group (p=0.0001), and almost all individuals (90.9%) with TMD had a high level of stress (p=0.0082). These findings suggest that there is positive association between TMD and environmental stress.