Editorial: Methods and applications in Alzheimer's disease and related dementias

Alzheimer's disease (AD) is the most prevalent neurodegenerative disease worldwide. Currently, AD diagnosis is based on a multidimensional approach that involves clinical, neuropathological examination and evaluation of biomarkers. The burden of AD is further exacerbated by the fact that brain damage actually begins to develop several years before the diagnosis or even mild cognitive impairment (MCI) is observed (Long and Holtzman, 2019). Therefore, developing early and accurate diagnostic methods is urgently needed. The present Research Topic aims to highlight the latest experimental techniques and methods used to investigate fundamental questions in Alzheimer's disease and related dementias, from integrative functions to molecular and potential therapeutics.

The challenge of the next decade: Screening of large populations with innovative approaches
The estimated total healthcare costs for the treatment of Alzheimer's disease in 2020 is estimated at $305 billion, with the cost expected to increase to more than $1 trillion as the population ages (Wong, 2020). It is proposed that early diagnosis and intervention are effective ways to reduce the burden of AD. The study by Ren et al., demonstrates the benefits of a screening program for AD in mainland China and debates the cost-effectiveness of implementing such programs for the health care system. Their report established increased health benefits and reduce the incidence of severe AD and death.

Diagnosis of AD is typically based on cognitive decline based on scores that can be affected by subjective factors. Machine learning techniques could be an attractive alternative for eliminating such biases and tracking cognitive decline and identifying risk factors in the progression of AD. Based on data from a large population-based longitudinal survey of elderly Chinese, Wang et al., were able to build a prediction model with machine learning algorithms to early identify the risk for cognitive impairment. This study demonstrates that risk-predictive models may serve in the future as a valuable tool in preventive health care.

Developments in magnetic resonance imaging (MRI), and other methods have resulted in the widespread use of brain imaging as a tool for AD diagnosis. Recently, the definition of new variables such as AD resemblance atrophy index (AD-RAI) has been introduced as a novel machine-learning-based brain atrophy biomarker for AD diagnosis (Mai et al., 2021). In contrast with single-brain regional biomarkers, the AD-RAI summarizes atrophy across multiple brain regions affected by AD. To date, the sample sizes to determine the potential clinical utility of the AD-RAI have been relatively small. Using the Minimal Interval Resonance Imaging in Alzheimer's disease (MIRIAD) database He et al., determined the same-day repeatability and 2-week reproducibility of AD-RAI and further validated its use in AD classification and prediction in a longitudinal setting.

One of the main challenges in the imaging field is regarding the reproducibility of samples acquired across multiple centers due to variety in scanner types and data acquisition protocols. A harmonization technique based on the linear mixed effect (LME) model is introduced by Kim et al., to overcome this issue. Using MRI cortical thickness measurements obtained from multiple centers, the authors showed that the score calculated by the LME method effectively compensates for the center effect across multiple datasets.