Subspace corrected relevance learning with application in neuroimaging.

In machine learning, data often comes from different sources, but combining them can introduce extraneous variation that affects both generalization and interpretability. For example, we investigate the classification of neurodegenerative diseases using FDG-PET data collected from multiple neuroimaging centers. However, data collected at different centers introduces unwanted variation due to differences in scanners, scanning protocols, and processing methods. To address this issue, we propose a two-step approach to limit the influence of center-dependent variation on the classification of healthy controls and early vs. late-stage Parkinson's disease patients. First, we train a Generalized Matrix Learning Vector Quantization (GMLVQ) model on healthy control data to identify a "relevance space" that distinguishes between centers. Second, we use this space to construct a correction matrix that restricts a second GMLVQ system's training on the diagnostic problem. We evaluate the effectiveness of this approach on the real-world multi-center datasets and simulated artificial dataset. Our results demonstrate that the approach produces machine learning systems with reduced bias - being more specific due to eliminating information related to center differences during the training process - and more informative relevance profiles that can be interpreted by medical experts. This method can be adapted to similar problems outside the neuroimaging domain, as long as an appropriate "relevance space" can be identified to construct the correction matrix.

PREvalence Study on Surgical COnditions (PRESSCO) 2020: A Population-Based Cross-Sectional Countrywide Survey on Surgical Conditions in Post-Ebola Outbreak Sierra Leone.

BACKGROUND: Understanding the burden of diseases requiring surgical care at national levels is essential to advance universal health coverage. The PREvalence Study on Surgical COnditions (PRESSCO) 2020 is a cross-sectional household survey to estimate the prevalence of physical conditions needing surgical consultation, to investigate healthcare-seeking behavior, and to assess changes from before the West African Ebola epidemic. METHODS: This study (ISRCTN: 12353489) was built upon the Surgeons Overseas Surgical Needs Assessment (SOSAS) tool, including expansions. Seventy-five enumeration areas from 9671 nationwide clusters were sampled proportional to population size. In each cluster, 25 households were randomly assigned and visited. Need for surgical consultations was based on verbal responses and physical examination of selected household members. RESULTS: A total of 3,618 individuals from 1,854 households were surveyed. Compared to 2012, the prevalence of individuals reporting one or more relevant physical conditions was reduced from 25 to 6.2% (95% CI 5.4-7.0%) of the population. One-in-five conditions rendered respondents unemployed, disabled, or stigmatized. Adult males were predominantly prone to untreated surgical conditions (9.7 vs. 5.9% women; p < 0.001). Financial constraints were the predominant reason for not seeking care. Among those seeking professional health care, 86.7% underwent surgery. CONCLUSION: PRESSCO 2020 is the first surgical needs household survey which compares against earlier study data. Despite the 2013-2016 Ebola outbreak, which profoundly disrupted the national healthcare system, a substantial reduction in reported surgical conditions was observed. Compared to one-time measurements, repeated household surveys yield finer granular data on the characteristics and situations of populations in need of surgical treatment.