A Regression Framework for Predicting Cognitive Decline in Frontotemporal Dementia using Recurrent Neural Networks.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Prognostics

Background:

• Frontotemporal dementia (FTD) is a progressive neurodegenerative disorder characterized by personality, behavioral, language, and executive function changes.
• FTD subtypes include behavioral variant FTD, non-fluent variant primary progressive aphasia, and semantic variant primary progressive aphasia.
• Early detection and understanding FTD progression are crucial, especially given its typical onset between ages 40-65.

Purpose of the Study:

• To forecast future cognitive status in individuals with FTD using longitudinal neuropsychological test scores.
• To evaluate the efficacy of an Encoder-Decoder Long-Short-Term-Memory (ED-LSTM) model for predicting FTD marker progression.
• To establish a method for early identification and prognosis of cognitive decline in FTD patients.

Main Methods:

• A regression framework utilizing an ED-LSTM model was applied to longitudinal data from the Frontotemporal Lobar Degeneration Neuroimaging Initiative (FTLDNI/NIFD).
• Data included neuropsychological test scores from 288 participants across 918 instances.
• The ED-LSTM model's performance was compared against standard LSTM and Simple RNN models using mean absolute error and root mean square error.

Main Results:

• The proposed ED-LSTM model demonstrated superior performance in forecasting FTD markers (cognitive scores) over a four-year period.
• The model achieved better accuracy metrics (mean absolute error and root mean square error) compared to baseline recurrent neural network models.
• This study is the first to comprehensively explore four-year forecasting of individual FTD markers.

Conclusions:

• The ED-LSTM model offers a promising approach for predicting cognitive decline in FTD.
• Accurate forecasting of FTD progression can significantly aid in early diagnosis and personalized treatment strategies.
• This research contributes to improving patient outcomes and the overall management of frontotemporal dementia.