Direct explanations for the development and use of a multi-layer perceptron network that classifies low-back-pain patients.

Using a new method published by the first author, this article shows how direct explanations can be provided to interpret the classification of any input case by a standard multilayer perceptron (MLP) network. The method is demonstrated for a real-world MLP that classifies low-back-pain patients into three diagnostic classes. The application of the method leads to the discovery of a number of mis-diagnosed training and test cases and to the development of a more optimal low-back-pain MLP network.

Derivation of the multilayer perceptron weight constraints for direct network interpretation and knowledge discovery.

This paper examines the multilayer perceptron (MLP) network from a hidden layer decision region perspective and derives the output layer and hidden layer weight constraints that the network must satisfy in performing a general classification task. This provides a foundation for direct knowledge discovery from the MLP, using a new method published by the author, which finds the key inputs that the MLP uses to classify an input case. The knowledge that the MLP network learns from the training examples is represented as ranked data relationships and induced rules, which can be used to validate the MLP network. The bounds of the network knowledge are established in the n-dimensional input space and a measure of the limit of the MLP network knowledge is proposed. An algorithm is presented for the calculation of the maximum number of hidden layer decision regions in the MLP input space.

Diagnosing ADHD (predominantly inattentive and combined type subtypes): discriminant validity of the behavior assessment system for children and the achenbach parent and teacher rating scales.

Compared the effectiveness of discriminating attention deficit/hyperactivity disorder (ADHD) subtypes using the Parent Rating Scale (PRS) and Teacher Rating Scale (TRS) of the Behavior Assessment System for Children (BASC) and the Parent Report Form and Teacher Report Form (TRF) of the Achenbach Child Behavior Checklist (CBCL). To determine the extent to which these scales measured similar behaviors, Pearson Product-Moment Correlations were computed for the parent scales (PRS and CBCL) and for the teacher scales (TRS and TRF). Results indicated that correlations were significant for a number of scales. Discriminant analysis does not suggest a strong advantage of either measure in differentiating children with ADHD from those who do not meet criteria for ADHD, except for the BASC TRS which has better predictive ability for children who do not meet ADHD criteria. For subtypes of ADHD, and specifically the ADHD: Predominantly Inattentive subtype, however, results would favor the use of the BASC PRS and TRS.

Decreased prevalence of Chlamydia trachomatis infection associated with a selective screening program in family planning clinics in Wisconsin.

The effectiveness of selective screening for control of Chlamydia trachomatis is unknown. In 1986, a statewide screening program began in family planning clinics in Wisconsin after the prevalence of infection among women was found to be 10.7% in four nonurban clinics and 13.7% in an urban Milwaukee clinic. In 1990, endocervical specimens were obtained from 1,757 women attending these same clinics; 5.2% of women in the non-urban clinics and 6.9% in the Milwaukee clinic tested positive for C. trachomatis. Prevalence of infection had decreased similarly (by 53% overall) in both high- and low-risk groups in all five clinics. Although reported condom use increased from 16% to 31%, most other demographic and behavioral risk factors for infection did not significantly change; in contrast, the prevalence of clinical signs of infection decreased. The percentage of infections identified by selective screening criteria decreased from 77% to 55%. Selective screening and attendant activities, as well as an increase in condom use, were associated with a decrease in prevalence of C. trachomatis infection in this population.

Chlamydia trachomatis infection in women attending urban midwestern family planning and community health clinics: risk factors, selective screening, and evaluation of non-culture techniques.

To determine prevalence and risk factors for endocervical Chlamydia trachomatis infection in an urban midwestern population and to evaluate two non-culture direct tests for C. trachomatis, we studied 849 women attending two family planning clinics and a community health clinic in Milwaukee, Wisconsin. Adequate endocervical specimens were obtained from 751 women for chlamydial isolation in tissue culture and antigen tests using direct fluorescence (DFA) and enzyme immunoassay (EIA); 93 (12.4%) patients had cultures positive for C. trachomatis. Compared to culture, the DFA test had a 77.4% sensitivity, 96.8% specificity, and a predictive value positive (PVP) of 77%. For the EIA, these values were 83.9%, 97.0%, and 80%, respectively. No single historical, clinical, or laboratory variable, including the previously described cervicitis index and specific cytologic findings on Pap smear, had sufficient predictive value to be used as the only criterion for selective screening in this population. Criteria for selective screening were proposed that would result in screening 43% of patients and would identify 71% of infections. PVP of both non-culture tests was 89% in persons identified by these criteria to be at increased risk of C. trachomatis infection.

Chlamydia: management in the physician's office.

Chlamydia trachomatis infections constitute the most prevalent bacterial sexually transmitted disease (STD) in Wisconsin. In 1987, chlamydia became Wisconsin's most frequently reported STD. Several significant clinical syndromes have been associated with chlamydia infection. Prevention and control of STD needs to become a basic part of primary care within the private medical setting. Determination of the need to test a patient for chlamydia infection cannot be based solely on the presence of signs or symptoms. Physicians in private practice need to identify high-risk patients and selectively screen high-risk men and women, rapidly initiate treatment of chlamydia-infected patients and their sexual partners, and work closely with public health personnel as part of disease intervention.

Selective screening for Chlamydia trachomatis infection in nonurban family planning clinics in Wisconsin.

To determine the prevalence of, and identify risk factors for, Chlamydia trachomatis infection, we studied 380 women attending four Wisconsin family planning clinics in October 1985. The patients completed self-administered sexual history questionnaires, were examined by nurse clinicians and had specimens taken for direct fluorescent antibody (DFA) testing for C. trachomatis. Of 335 women with adequate specimens, 10.7 percent had positive DFA tests. Selective screening criteria were developed based on the following risk factors for C. trachomatis: Age less than 20 years and recent exposure to either a new sexual partner or a partner with more than one partner; a partner with symptoms of urethritis; a diagnosis of cervicitis; and inflammatory changes on Pap smear. Thirty-six percent of patients met one or more of these screening criteria, and the criteria had a sensitivity of 72 percent.

PIP: In October 1985, clinicians took an endocervical sample from all women (at least 335) who attended family planning clinics in Wausau, Manitowoc, Eau Claire, and Superior in Wisconsin to determine prevalence and risk factors for chlamydial infection among rural midwestern women. Laboratory staff used the direct fluorescent antibody test for Chlamydia trachomatis. Each patient completed a questionnaire on her clinical and sexual history. 36 (10.7%) had a chlamydial infection. Women who came to the clinic because of a possible sexually transmitted disease or gynecologic infection (11.4%) were just as likely to have a chlamydial infection as were those who came for a routine exam (11.7%). Thus a women's reason for visiting a clinic was not predictive of chlamydial infection. This finding indicated that clinicians should determine the presence of risk factors even in women who come for a routine exam. Further it also indicated that women who visited the clinic with genitourinary symptoms did not necessarily mean a chlamydial infection. Sexual exposure risk factors included a new sexual partner within 3 months before the exam (p=.05; relative risk [RR]=2), a partner with 1 partner within 3 months before the exam (p=.007; RR=3.2), and a partner with signs or symptoms of urethritis within 30 days before the exam (p=.0004; R=14.4). Of all women with C. trachomatis, 40% had mucopurulent discharge and 8.9% did not have it (p=.0001; RR=4.5). Moreover, 26.5% had a friable cervix compared to 8% without a friable cervix (p=.0001; Rr=3.3). Chlamydial infection was also significantly associated with inflammatory changes on the most recent Pap smear (p=.0001; Rr=5.2). Further women 20 years were 2.4 times more likely to have a chlamydial infection than those 20 years old (p=.02). 36% fulfilled at least 1 of the screening criteria. Criteria sensitivity was 72%.