Pattern recognition as a concept for multiple-choice questions in a national licensing exam.

BACKGROUND: Multiple-choice questions (MCQ) are still widely used in high stakes medical exams. We wanted to examine whether and to what extent a national licensing exam uses the concept of pattern recognition to test applied clinical knowledge. METHODS: We categorized all 4,134 German National medical licensing exam questions between October 2006 and October 2012 by discipline, year, and type. We analyzed questions from the four largest disciplines: internal medicine (n = 931), neurology (n = 305), pediatrics (n = 281), and surgery (n = 233), with respect to the following question types: knowledge questions (KQ), pattern recognition questions (PRQ), inverse PRQ (IPRQ), and pseudo PRQ (PPRQ). RESULTS: A total 51.1% of all questions were of a higher taxonomical order (PRQ and IPRQ) with a significant decrease in the percentage of these questions (p <0.001) from 2006 (61.5%) to 2012 (41.6%). The proportion of PRQs and IPRQs was significantly lower (p <0.001) in internal medicine and surgery, compared to neurology and pediatrics. PRQs were mostly used in questions about diagnoses (71.7%). A significantly higher (p <0.05) percentage of PR/therapy questions was found for internal medicine compared with neurology and pediatrics. CONCLUSION: The concept of pattern recognition is used with different priorities and to various extents by the different disciplines in a high stakes exam to test applied clinical knowledge. Being aware of this concept may aid in the design and balance of MCQs in an exam with respect to testing clinical reasoning as a desired skill at the threshold of postgraduate medical education.

A comparative study of pattern recognition classifiers to predict physical activities using smartphones and wearable body sensors.

This paper presents a wireless body area network platform that performs physical activities recognition using accelerometers, biosignals and smartphones. Multiple classifiers and sensor combinations were examined to identify the classifier with the best recognition performance for the static and dynamic activities. The Functional Trees classifier proved to provide the best results among the classifiers evaluated (Naive Bayes, Bayesian Networks, Support Vector Machines and Decision Trees [C4.5, Random Forest]) and was used to train the model which was implemented for the real time activity recognition on the smartphone. The identified patterns of daily physical activities were used to examine conformance with medical advice, regarding physical activity guidelines. An algorithm based on Skip Chain Conditional Random Fields, received as inputs the recognized activities and data retrieved from the GPS receiver of the smartphone to develop dynamic daily patterns that enhance prediction results. The presented platform can be extended to be used in the prevention of short-term complications of metabolic diseases such as diabetes.

Lexical tone identification and consonant recognition in acoustic simulations of cochlear implants.

CONCLUSION: Cochlear implant (CI) recipients' performance of lexical tone identification and consonant recognition can be enhanced by providing greater spectral details. OBJECTIVE: To evaluate the effects of increasing the number of total spectral channels on the lexical tone identification and consonant recognition by normally hearing listeners who are native speakers of Mandarin Chinese. SUBJECTS AND METHODS: Lexical tone identification and consonant recognition were measured in 15 Mandarin-speaking, normal-hearing (NH) listeners with varied numbers of total spectral channels (i.e. 4, 6, 8, 10, 12, 16, 20, and 24), using acoustic simulations of CIs. RESULTS: The group of NH listeners' performance of lexical tone identification ranged from 44.53% to 66.60% with 4-24 spectral channels. The performance of tone identification between channels 4 and 16 remained similar; between channels 16 and 20 performance improved significantly. As regards consonant recognition, the NH listeners' overall accuracy ranged from 73.17% to 95.33% with 4-24 channels. Steady improvement in consonant recognition accuracy was observed as a function of increasing the spectral channels. With about 12-16 spectral channels, the NH listeners' overall accuracy in consonant recognition began to be comparable to their accuracy with the unprocessed stimuli.