Normative Diffusion Autoencoders: Application to Amyotrophic Lateral Sclerosis

Ayodeji Ijishakin,Adamos Hadjasavilou,Ahmed Abdulaal,Nina Montana-Brown,Florence Townend,Edoardo Spinelli,Massimo Fillipi,Federica Agosta,James Cole,Andrea Malaspina

2024-07-19

Abstract:Predicting survival in Amyotrophic Lateral Sclerosis (ALS) is a challenging task. Magnetic resonance imaging (MRI) data provide in vivo insight into brain health, but the low prevalence of the condition and resultant data scarcity limit training set sizes for prediction models. Survival models are further hindered by the subtle and often highly localised profile of ALS-related neurodegeneration. Normative models present a solution as they increase statistical power by leveraging large healthy cohorts. Separately, diffusion models excel in capturing the semantics embedded within images including subtle signs of accelerated brain ageing, which may help predict survival in ALS. Here, we combine the benefits of generative and normative modelling by introducing the normative diffusion autoencoder framework. To our knowledge, this is the first use of normative modelling within a diffusion autoencoder, as well as the first application of normative modelling to ALS. Our approach outperforms generative and non-generative normative modelling benchmarks in ALS prognostication, demonstrating enhanced predictive accuracy in the context of ALS survival prediction and normative modelling in general.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict the survival time of patients with Amyotrophic Lateral Sclerosis (ALS). ALS is a rare and incurable neurodegenerative disease, with a global incidence rate of 4.36 per 100,000 people. Due to the low incidence rate of ALS and data scarcity, as well as the relatively short survival time of patients (the median survival time is 3 years), it is particularly difficult to predict the survival time of ALS patients.

Specifically, this research faces the following challenges:
1. **Data Scarcity**: The MRI data of ALS patients are limited, resulting in a relatively small dataset for training the prediction model.
2. **Subtle Neurodegenerative Changes**: The neurodegenerative changes related to ALS are usually very subtle and highly localized, and are difficult to detect in neuroimaging data.
3. **Limitations of Existing Methods**: Traditional survival models are not effective in handling high - dimensional and complex neuroimaging data, especially for capturing subtle neuroanatomical changes.

To solve these problems, researchers introduced a new framework - Normative Diffusion Autoencoders. This framework combines the advantages of generative models and normative models to improve prediction accuracy in the following ways:
- **Normative Model**: Utilize the data of large - scale healthy cohorts to increase statistical power, thereby improving the sensitivity to abnormal data points.
- **Diffusion Model**: Be good at capturing semantic information in images, including subtle signs of accelerated brain aging, which is helpful for predicting the survival time of ALS patients.

Specifically, this research proposed a Hierarchical Diffusion Autoencoder, and calculated the cosine similarity between the latent representation of ALS patients' MRI and the average latent representation of the healthy cohort as a feature for survival time prediction. Experimental results show that this method is superior to the existing generative and non - generative normative model benchmarks in ALS survival prediction.

In conclusion, this paper aims to overcome the problems of data scarcity and subtle nerve change detection in ALS survival prediction by introducing the Normative Diffusion Autoencoder framework, thereby providing more accurate survival time prediction.