We currently observe a disconcerting phenomenon in machine learning studies in psychiatry: While we would expect larger samples to yield better results due to the availability of more data, larger machine learning studies consistently show much weaker performance than the numerous small-scale studies. Here, we systematically investigated this effect focusing on one of the most heavily studied questions in the field, namely the classification of patients suffering from Major Depressive Disorder (MDD) and healthy controls based on neuroimaging data. Drawing upon structural MRI data from a balanced sample of N = 1868 MDD patients and healthy controls from our recent international Predictive Analytics Competition (PAC), we first trained and tested a classification model on the full dataset which yielded an accuracy of 61%. Next, we mimicked the process by which researchers would draw samples of various sizes (N = 4 to N = 150) from the population and showed a strong risk of misestimation. Specifically, for small sample sizes (N = 20), we observe accuracies of up to 95%. For medium sample sizes (N = 100) accuracies up to 75% were found. Importantly, further investigation showed that sufficiently large test sets effectively protect against performance misestimation whereas larger datasets per se do not. While these results question the validity of a substantial part of the current literature, we outline the relatively low-cost remedy of larger test sets, which is readily available in most cases.

Thas work focuses o n creatang a framework f o r objectavely evaluatang the performance of range amage segmentataon algorathms. T h e algorathms are evaluated an terms of correct segmentataon, over-and under-segmentataon, massed and noase regaons. A set of amages wath ground truth was created f o r thas work. T h e amages were captured usang a structured laght scanner. Images used an the evaluataon contaan planar, spheracal, cylandracal, torozdal and conacal surface patches. T h e dafferent surface patches an each zmage were manually adentafied to establash ground truth f o r performance evaluataon. T w o segmentataon algorzthms f r o m the laterature are compared.

No abstract

No abstract

No abstract

No abstract

No abstract

No abstract