Why is Real-World Visual Object Recognition Hard?

Progress in understanding the brain mechanisms underlying vision requires the construction of computational models that not only emulate the brain's anatomy and physiology, but ultimately match its performance on visual tasks. In recent years, “natural” images have become popular in the study of vision and have been used to show apparently impressive progress in building such models. Here, we challenge the use of uncontrolled “natural” images in guiding that progress. In particular, we show that a simple V1-like model—a neuroscientist's “null” model, which should perform poorly at real-world visual object recognition tasks—outperforms state-of-the-art object recognition systems (biologically inspired and otherwise) on a standard, ostensibly natural image recognition test. As a counterpoint, we designed a “simpler” recognition test to better span the real-world variation in object pose, position, and scale, and we show that this test correctly exposes the inadequacy of the V1-like model. Taken together, these results demonstrate that tests based on uncontrolled natural images can be seriously misleading, potentially guiding progress in the wrong direction. Instead, we reexamine what it means for images to be natural and argue for a renewed focus on the core problem of object recognition—real-world image variation.     

High-Resolution Mapping of a Genetic Locus Regulating Preferential Carbohydrate Intake,Total Kilocalories,and Food Volume on Mouse Chromosome 17

The specific genes regulating the quantitative variation in macronutrient preference and food intake are virtually unknown. We fine mapped a previously identified mouse chromosome 17 region harboring quantitative trait loci (QTL) with large effects on preferential macronutrient intake-carbohydrate (Mnic1), total kilcalories (Kcal2), and total food volume (Tfv1) using interval-specific strains. These loci were isolated in the [C57BL/6J.CAST/EiJ-17.1-(D17Mit19-D17Mit50); B6.CAST-17.1] strain, possessing a ∼40.1 Mb region of CAST DNA on the B6 genome. In a macronutrient selection paradigm, the B6.CAST-17.1 subcongenic mice eat 30% more calories from the carbohydrate-rich diet, ∼10% more total calories, and ∼9% more total food volume per body weight. In the current study, a cross between carbohydrate-preferring B6.CAST-17.1 and fat-preferring, inbred B6 mice was used to generate a subcongenic-derived F₂ mapping population; genotypes were determined using a high-density, custom SNP panel. Genetic linkage analysis substantially reduced the 95% confidence interval for Mnic1 (encompassing Kcal2 and Tfv1) from 40.1 to 29.5 Mb and more precisely established its boundaries. Notably, no genetic linkage for self-selected fat intake was detected, underscoring the carbohydrate-specific effect of this locus. A second key finding was the separation of two energy balance QTLs: Mnic1/Kcal2/Tfv1 for food intake and a newly discovered locus regulating short term body weight gain. The Mnic1/Kcal2/Tfv1 QTL was further de-limited to 19.0 Mb, based on the absence of nutrient intake phenotypes in subcongenic HQ17IIa mice. Analyses of available sequence data and gene ontologies, along with comprehensive expression profiling in the hypothalamus of non-recombinant, cast/cast and b6/b6 F₂ controls, focused our attention on candidates within the QTL interval. Zfp811, Zfp870, and Btnl6 showed differential expression and also contain stop codons, but have no known biology related to food intake regulation. The genes Decr2, Ppard and Agapt1 are more appealing candidates because of their involvement in lipid metabolism and down-regulation in carbohydrate-preferring animals.     

Reduced susceptibility to ventricular tachyarrhythmias in rats selectively bred for high aerobic capacity

Reperfusion after a brief period of cardiac ischemia can lead to potentially lethal arrhythmias. Human epidemiological studies and experimental work with animals indicate that regular physical activity is associated with reductions in cardiovascular disease (CVD) risk factors and sudden cardiac death. Similarly, artificial selection of rats for high aerobic treadmill-running capacity (high-capacity runners; HCR) has been shown to reduce CVD risk factors relative to rats selected as low-capacity runners (LCR). Therefore, we tested the hypothesis that HCR, relative to LCR rats, would be less susceptible to ischemia-reperfusion-mediated ventricular tachyarrhythmias. To test this hypothesis, we measured the susceptibility to ventricular tachyarrhythmias produced by 3 min of occlusion and reperfusion of the left main coronary artery in conscious LCR and HCR rats. Results document a significantly lower incidence of ventricular tachyarrhythmias in HCR (3 of 11, 27.3%) relative to LCR (6 of 7, 85.6%) rats. The decreased susceptibility to tachyarrhythmias in HCR rats was associated with a reduced cardiac metabolic demand during ischemia (lower rate-pressure product and ST segment elevation) as well as a wider range for the autonomic control of heart rate. The HCR and LCR represent a unique substrate for evaluation of the mechanisms underlying ischemia-mediated cardiac arrhythmogenesis.     

Postexercise alpha-adrenergic receptor hyporesponsiveness in hypertensive rats is due to nitric oxide

We tested the hypothesis that a single bout of dynamic exercise produces a postexercise hypotension (PEH) and alpha(1)-adrenergic receptor hyporesponsiveness in spontaneously hypertensive rats (SHR). The postexercise alpha(1)-adrenergic receptor hyporesponsiveness is due to an enhanced buffering of vasoconstriction by nitric oxide. Male (n = 8) and female (n = 5) SHR were instrumented with a Doppler ultrasonic flow probe around the femoral artery. Distal to the flow probe, a microrenathane catheter was inserted into a branch of the femoral artery for the infusion of the alpha(1)-adrenergic receptor agonist phenylephrine (PE). A microrenathane catheter was inserted into the descending aorta via the left common carotid artery for measurements of arterial pressure (AP) and heart rate. Dose-response curves to PE (3.8 x 10(-3) - 1.98 x 10(-2)microg/kHz) were generated before and after a single bout of dynamic exercise. Postexercise AP was reduced in male (13 +/- 3 mmHg) and female SHR (18 +/- 7 mmHg). Postexercise vasoconstrictor responses to PE were reduced in males due to an enhanced influence of nitric oxide. However, in females, postexercise vasoconstrictor responses to PE were not altered. Results suggest that nitric oxide- mediated alpha(1)-adrenergic receptor hyporesponsiveness contributes to PEH in male but not female SHR.     

Peer instruction improves performance on quizzes

Peer instruction is a cooperative-learning technique that promotes critical thinking, problem solving, and decision-making skills. Benson's think-pair-share and Mazur's peer-instruction techniques are simple cooperative exercises that promote student's participation in class and increase student's interaction with each other and with the instructor in a large classroom. We borrowed concepts from Benson and Mazur and applied these concepts to enhance student involvement during the respiratory component of the medical physiology class. The medical physiology class consisted of 256 first-year medical students. The peer-instruction technique was used for 10 classes. Each class of 50 min was divided into three or four short presentations of 12-20 min. Each presentation was followed by a one-question, multiple-choice quiz on the subject discussed. Questions ranged from simple recall to those testing complex intellectual activities. Students were given 1 min to think and to record their first answer. Subsequently, students were allowed 1 min to discuss their answers with their classmates and possibly correct their first response. The percentage of correct answers increased significantly (P < 0.05) after discussion for both recall and intellectual questions. These data demonstrate that pausing three to four times during a 50-min class to allow discussion of concepts enhanced the students level of understanding and ability to synthesize and integrate material.     

Who wants to be a physician? An educational tool for reviewing pulmonary physiology

Traditional review sessions are typically focused on instructor-based learning. However, experts in the field of higher education have long recommended teaching modalities that incorporate student-based active-learning strategies. Given this, we developed an educational game in pulmonary physiology for first-year medical students based loosely on the popular television game show Who Wants To Be A Millionaire. The purpose of our game, Who Wants To Be A Physician, was to provide students with an educational tool by which to review material previously presented in class. Our goal in designing this game was to encourage students to be active participants in their own learning process. The Who Wants To Be A Physician game was constructed in the form of a manual consisting of a bank of questions in various areas of pulmonary physiology: basic concepts, pulmonary mechanics, ventilation, pulmonary blood flow, pulmonary gas exchange, gas transport, and control of ventilation. Detailed answers are included in the manual to assist the instructor or player in comprehension of the material. In addition, an evaluation instrument was used to assess the effectiveness of this instructional tool in an academic setting. Specifically, the evaluation instrument addressed five major components, including goals and objectives, participation, content, components and organization, and summary and recommendations. Students responded positively to our game and the concept of active learning. Moreover, we are confident that this educational tool has enhanced the students' learning process and their ability to understand and retain information.     

Active learning of respiratory physiology improves performance on respiratory physiology examinations

Active involvement in the learning process has been suggested to enhance creative thinking, judgement, interpretation, and problem-solving skills. Therefore, educators are encouraged to create an active-learning environment by incorporating active-learning strategies into the class. However, there is very little documentation of the effectiveness of active-learning strategies. Furthermore, faculty are often reluctant to incorporate new strategies without documentation of the effectiveness of these strategies. To address this concern, we compared the performance of two individual classes on an identical respiratory physiology examination. One class was taught respiratory physiology using active-learning strategies. The other class was taught respiratory physiology using the traditional lecture format. The results document that students who learned using active-learning strategies did significantly better (P < 0.05) on the respiratory physiology examination than students who learned by the traditional lecture format (61 +/- 2.2 vs. 86 +/- 1.0). Thus, by actively involving students in the learning process, academic performance is enhanced.     

Mechanisms mediating NTS P2x receptor-evoked hypotension: cardiac output vs. total peripheral resistance

We have previously shown that P2x purinoceptor activation in the subpostremal nucleus tractus solitarius (NTS) produces dose-dependent decreases in mean arterial pressure (MAP), heart rate, efferent sympathetic nerve activity, and significant peripheral vasodilation. However, the relative roles of cardiac output (CO) and total peripheral resistance (TPR) in mediating this depressor response are unknown. Bradycardia does not necessarily result in decreased CO, because, with the greater filling time, stroke volume may increase such that CO may be unchanged. We measured changes in CO (via a chronically implanted flow probe on the ascending aorta) and MAP in alpha-chloralose- and urethane-anesthetized male Sprague-Dawley rats in response to microinjection of the selective P2x purinoceptor agonist alpha,beta-methylene ATP (25 and 100 pmol/50 nl) into the subpostremal NTS. TPR was calculated as MAP/CO. At the low dose of NTS P2x purinoceptor agonist, the reduction in MAP was primarily mediated by reductions in TPR (-31.3 +/- 3.3%), not CO (-8.7 +/- 1.7%). At the high dose, both CO (-34.4 +/- 6.6%) and TPR (-40.2 +/- 2.5%) contribute to the reduction in MAP. We conclude that the relative contribution of CO and TPR to the reduction in MAP evoked by NTS P2x purinoceptor activation is dependent on the extent of P2x purinoceptor activation.     

Feasibility of an Ingestible Sensor-Based System for Monitoring Adherence to Tuberculosis Therapy

Poor adherence to tuberculosis (TB) treatment hinders the individual’s recovery and threatens public health. Currently, directly observed therapy (DOT) is the standard of care; however, high sustaining costs limit its availability, creating a need for more practical adherence confirmation methods. Techniques such as video monitoring and devices to time-register the opening of pill bottles are unable to confirm actual medication ingestions. A novel approach developed by Proteus Digital Health, Inc. consists of an ingestible sensor and an on-body wearable sensor; together, they electronically confirm unique ingestions and record the date/time of the ingestion. A feasibility study using an early prototype was conducted in active TB patients to determine the system’s accuracy and safety in confirming co-ingestion of TB medications with sensors. Thirty patients completed 10 DOT visits and 1,080 co-ingestion events; the system showed 95^.0% (95% CI 93^.5–96^.2%) positive detection accuracy, defined as the number of detected sensors divided by the number of transmission capable sensors administered. The specificity was 99^.7% [95% CI 99^.2–99^.9%] based on three false signals recorded by receivers. The system’s identification accuracy, defined as the number of correctly identified ingestible sensors divided by the number of sensors detected, was 100%. Of 11 adverse events, four were deemed related or possibly related to the device; three mild skin rashes and one complaint of nausea. The system’s positive detection accuracy was not affected by the subjects’ Body Mass Index (p = 0^.7309). Study results suggest the system is capable of correctly identifying ingestible sensors with high accuracy, poses a low risk to users, and may have high patient acceptance. The system has the potential to confirm medication specific treatment compliance on a dose-by-dose basis. When coupled with mobile technology, the system could allow wirelessly observed therapy (WOT) for monitoring TB treatment as a replacement for DOT.