(Predictable) performance bias in unsupervised anomaly detection.

BACKGROUND: With the ever-increasing amount of medical imaging data, the demand for algorithms to assist clinicians has amplified. Unsupervised anomaly detection (UAD) models promise to aid in the crucial first step of disease detection. While previous studies have thoroughly explored fairness in supervised models in healthcare, for UAD, this has so far been unexplored. METHODS: In this study, we evaluated how dataset composition regarding subgroups manifests in disparate performance of UAD models along multiple protected variables on three large-scale publicly available chest X-ray datasets. Our experiments were validated using two state-of-the-art UAD models for medical images. Finally, we introduced subgroup-AUROC (sAUROC), which aids in quantifying fairness in machine learning. FINDINGS: Our experiments revealed empirical "fairness laws" (similar to "scaling laws" for Transformers) for training-dataset composition: Linear relationships between anomaly detection performance within a subpopulation and its representation in the training data. Our study further revealed performance disparities, even in the case of balanced training data, and compound effects that exacerbate the drop in performance for subjects associated with multiple adversely affected groups. INTERPRETATION: Our study quantified the disparate performance of UAD models against certain demographic subgroups. Importantly, we showed that this unfairness cannot be mitigated by balanced representation alone. Instead, the representation of some subgroups seems harder to learn by UAD models than that of others. The empirical "fairness laws" discovered in our study make disparate performance in UAD models easier to estimate and aid in determining the most desirable dataset composition. FUNDING: European Research Council Deep4MI.

Cognitive Fatigue Predicts Cognitive Failure in Multiple Sclerosis Patients and Healthy Controls: A Case-Control Study.

OBJECTIVE: Fatigue and cognitive deficits are frequent symptoms of multiple sclerosis (MS). However, the exact nature of their co-occurrence is not fully understood. We sought to determine the impact of cognitive and physical fatigue on subjective cognitive deficits in MS patients and healthy controls. METHODS: Self-reports of fatigue (FSMC), depression (CES-D), cognitive deficits (CFQ), and personality traits (NEO-FFI, ANPS) among 30 MS inpatients and 30 healthy controls were analyzed using hierarchical regression models. The frequency of cognitive mistakes was used as the dependent variable and the extent of cognitive and physical fatigue as the independent variable. RESULTS: Cognitive fatigue was the only unique and significant predictor of cognitive mistakes in both groups, explaining 13.3% of additional variance in the MS group after correcting for age, mood, and physical fatigue. Physical fatigue had no significant impact on cognitive mistakes. While age had an impact on cognitive mistakes and depression in healthy controls, this association was not significant in MS patients. Depression was significantly correlated with cognitive mistakes and cognitive fatigue in MS patients. CONCLUSIONS: The interplay of cognitive fatigue and subjective cognitive impairment can be generalized, with the exception of the variables of age and depression, which were shown to have differing impacts on cognitive mistakes in MS patients and healthy controls, respectively. Cognitive fatigue was linked to cognitive mistakes even after correcting for overlapping items in MS patients only. Future research should further investigate the link between cognitive fatigue and attention lapses in daily life by using various objective assessments.

Prenatal testosterone and stuttering.

BACKGROUND: The prevalence of stuttering is much higher in males compared to females. The biological underpinnings of this skewed sex-ratio is poorly understood, but it has often been speculated that sex hormones could play an important role. AIMS: The present study investigated a potential link between prenatal testosterone and stuttering. Here, an indirect indicator of prenatal testosterone levels, the Digit Ratio (2D:4D) of the hand, was used. As numerous studies have shown, hands with more "male" characteristics (putatively representing greater prenatal testosterone levels) are characterized by a longer ring finger compared to the index finger (represented as a lower 2D:4D ratio) in the general population. STUDY DESIGN, SUBJECTS, OUTCOME MEASURES: We searched for differences in the 2D:4D ratios between 38 persons who stutter and 36 persons who do not stutter. In a second step, we investigated potential links between the 2D:4D ratio and the multifaceted symptomatology of stuttering, as measured by the Overall Assessment of the Speaker's Experience of Stuttering (OASES), in a larger sample of 44 adults who stutter. RESULTS: In the first step, no significant differences in the 2D:4D were observed between individuals who stutter and individuals who do not stutter. In the second step, 2D:4D correlated negatively with higher scores of the OASES (representing higher negative experiences due to stuttering), and this effect was more pronounced for female persons who stutter. CONCLUSIONS: The findings indicate for the first time that prenatal testosterone may influence individual differences in psychosocial impact of this speech disorder.