ICAM-reg: Interpretable Classification and Regression with Feature Attribution for Mapping Neurological Phenotypes in Individual Scans

Abstract

An important goal of medical imaging is to be able to precisely detect patterns of disease specific to individual scans; however, this is challenged in brain imaging by the degree of heterogeneity of shape and appearance. Traditional methods, based on image registration to a global template, historically fail to detect variable features of disease, as they utilise population-based analyses, suited primarily to studying group-average effects. In this paper we therefore take advantage of recent developments in generative deep learning to develop a method for simultaneous classification, or regression, and feature attribution (FA). Specifically, we explore the use of a VAE-GAN translation network called ICAM, to explicitly disentangle class relevant features from background confounds for improved interpretability and regression of neurological phenotypes. We validate our method on the tasks of Mini-Mental State Examination (MMSE) cognitive test score prediction for the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, as well as brain age prediction, for both neurodevelopment and neurodegeneration, using the developing Human Connectome Project (dHCP) and UK Biobank datasets. We show that the generated FA maps can be used to explain outlier predictions and demonstrate that the inclusion of a regression module improves the disentanglement of the latent space. Our code is freely available on Github https: //github.com/CherBass/ICAM.

A Deep Generative Model of Neonatal Cortical Surface Development

“…One class of methods utilise variational encodings such as variational autoencoders (VAEs) [18], and VAE-GANs [22], although the authors found that attempts to adapt these methods to surface domains were unsuccessful, with models unable to capture individual cortical folding variation and collapsing to group averages. A more powerful example of this applied to volumetric cortical data is the iCAM architecture [3], which encodes separate variational disentangled spaces for content and age, which may be more amenable to adaptation with gDL. There are further alternatives that utilise direct conditioning on latent variables such as conditional VAEs and conditional GANs, but these too are more commonly associated with conditioning on classes, not continuous variables.…”

“…Deep Generative modelling presents enormous opportunities for medical imaging analysis: from image segmentation [13,9,40,8], registration [39], motion correction and denoising [30,42,23,1], to anomaly detection [14,45,4] and the development of clinically interpretable models of disease progression [2,3,25]. Image-toimage translation is a type of generative modelling problem where images are transformed across domains, in a way that preserves their content.…”

The effect of choosing optimizer algorithms to improve computer vision tasks: a comparative study

“…In the alignment stage, a layer cannot learn unless its upper layers are roughly aligned. Bass et al [10], who developed the human connectome project, used UK Biobank datasets. Their method was validated via Mini-Mental State Examination cognitive test score prediction for Alzheimer's disease The neuroimaging initiative cohort and brain age prediction for both neurodevelopment and neurodegeneration were considered.…”

“…These measures can be computed on the basis of the following Eqs. (10)(11)(12)(13)(14)(15)(16)(17)(18)(19) [72]:…”

Towards better Interpretable and Generalizable AD detection using Collective Artificial Intelligence

“…Thus, this kind of method seems to be more valuable to understand the model decision. However, to the best of our knowledge, there is currently few interpretable methods for AD classification, with our definition [38,39].…”