A predictive model using the mesoscopic architecture of the living brain to detect Alzheimer's disease.

Background: Alzheimer's disease, the most common cause of dementia, causes a progressive and irreversible deterioration of cognition that can sometimes be difficult to diagnose, leading to suboptimal patient care.
Methods: We developed a predictive model that computes multi-regional statistical morpho-functional mesoscopic traits from T1-weighted MRI scans, with or without cognitive scores. For each patient, a biomarker called "Alzheimer's Predictive Vector" (ApV) was derived using a two-stage least absolute shrinkage and selection operator (LASSO).
Results: The ApV reliably discriminates between people with (ADrp) and without (nADrp) Alzheimer's related pathologies (98% and 81% accuracy between ADrp - including the early form, mild cognitive impairment - and nADrp in internal and external hold-out test sets, respectively), without any a priori assumptions or need for neuroradiology reads. The new test is superior to standard hippocampal atrophy (26% accuracy) and cerebrospinal fluid beta amyloid measure (62% accuracy). A multiparametric analysis compared DTI-MRI derived fractional anisotropy, whose readout of neuronal loss agrees with ADrp phenotype, and SNPrs2075650 is significantly altered in patients with ADrp-like phenotype. Conclusions: This new data analytic method demonstrates potential for increasing accuracy of Alzheimer diagnosis.

Introduction

Alzheimer’s disease (AD) is the most common cause of dementia worldwide and is characterised by progressive cognitive impairment and brain atrophy[1]. The disease is characterised by several events. The National Institute on Aging and Alzheimer’s Association has proposed a classification system to categorise individuals based on biomarker evidence of pathology. This is called the ATN classification system and is used to rate people for the presence of cerebrospinal fluid β-amyloid (CSF Aβ or amyloid positron emission tomography (PET): 'A'), hyperphosphorylated τ (CSF pτ or τ PET: 'T'), and neurodegeneration (atrophy on structural magnetic resonance imaging (MRI), FDG) PET, or CSF total τ: 'N'), resulting in eight possible biomarker combinations[2]. Furthermore, a recent report on the involvement of microglial activation in the spread of τ tangles over the neocortex in AD suggests an additional inflammation biomarker for AD[3]. The most consistent structural imaging finding in AD is the reduced hippocampal volume[4], but this is arguably not the most specific structural biomarker as AD frequently presents with non-amnestic symptoms with initial involvement of extra-temporal regions of the brain[5]. Furthermore, the reduced hippocampal volume has been found in many other neuropsychiatric conditions including schizophrenia[6], depression[7] and hippocampal sclerosis[8] as well as the recently described limbic-predominant age-related TDP-43 encephalopathy[9]. Together with the hippocampal volume, Aβ(1–42), phosphorylated τ (pτ), and total τ (τ) CSF biomarkers have been shown to discriminate patients with AD from healthy controls[10]. However, their introduction into clinical practice is limited by considerable variability between laboratories and assay batches[10]. Similarly, blood-based biomarkers, which are eagerly awaited to address issues related to the invasiveness and high cost of CSF-based ones, often stall in the early stages because of a disconnect between academia, where biomarkers are identified, and industry, where they should be developed and commercially distributed[11].

In these last 40 years, improved computational power and storage capacity have led to numerous advances in developing non-invasive and low-cost structural biomarkers for AD that combine neuroimaging approaches, in particular structural MRI[12], with machine learning. This approach involves the acquisition of image data, the segmentation of the region of interest (ROI), feature extraction and selection for classification/prediction. Critically, features extracted from radiological images are able to reveal useful new biology[13,14] hidden to the clinician’s eye[15]—at a mesoscopic scale. For example, the mesoscopic architecture of entire tumours can reveal stromal phenotype or immune context, with strong prognostic or predictive utility[16,17]. In a radiomics analysis, the extracted features represent statistical morpho-functional traits of intensity, shape, texture, scale, grey level co-occurrence matrix (GLCM), grey level run-length matrix (RLM), grey level size zone matrix (GLSZM), neighbourhood grey tone difference matrix (NGTDM) and neighbourhood grey level dependence matrix (NGLDM)[18]. A number of studies have shown texture differences between AD patients and healthy controls (HC) in structures such as the hippocampus, corpus callosum, and thalamus[19,20]. Supplementary Data 1 summarises the results and methods of the most cited papers published in the last 5 years on the classification of AD and AD-related mild cognitive impairment (MCI) patients using multimodal features. Zhang et al.[21] for instance used a single-hidden-layer neural network and predator-prey particle swarm optimisation algorithm to classify HC from AD patients. They extracted texture features from one selected axial slice of a T1-weighted (T1w) MRI scan and obtained 93% accuracy in an internal test set. Similarly, Sorensen et al.[22], with a linear discriminant analysis extracted cortical thickness measurements, volumetric measurements and hippocampal volume, shape and texture features and reached from a T1w MRI scan with 63% accuracy. With the integration of genetic and cerebrospinal fluid biomarkers, Tong et al.[23] reached a 0.78 area under the curve (AUC) in the discrimination between HC and people with an AD-related mild cognitive impairment, thus pushing the technology towards earlier detection. They used a non-linear graph fusion method to reduce the number of volumetric features extracted from T1w MRI, intensity features extracted from PET data, three CSF measures and one genetic categorical feature. An improved performance was obtained with the view-aligned hypergraph learning approach used by Lin et al.[24]. They obtained 93, 90, 80 and 79% accuracies in the discrimination between HC and AD patients, HC and progressive MCI, HC and MCI, and stable and progressive MCI patients, respectively. In aggregate, when all patients, including control, prodromal forms of AD and AD are combined, most methods reach lower accuracy values. Of note, in most studies, models were trained and tested on an internal dataset only (Supplementary Data 1).

This current study proposes a method able to characterise early and later forms of Alzheimer’s disease with the extraction from a T1w MRI sequence of 29,520 statistical morpho-functional traits distributed over a multi-regional brain mask obtained with an automatic segmentation. Healthy brain and diseases unrelated to AD pathology, including Parkinson’s disease and frontotemporal dementia have been combined for the development of a set of tools able to reveal the mesoscopic architecture unique to AD.

Methods

The study workflow is summarised in Fig. 1. The analysis of baseline age-matched T1w MRI images consisted of a two-step combined approach with and without the additional information given by cognitive scores and CSF-based biomarkers. The model was trained on 1.5 T T1w MRI scans obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). After stratified randomisation, 70% of data were used for training and 30% for validation (robustness test shown in Supplementary Fig. 1). The control group (nADrp) included healthy controls, patients with frontotemporal dementia and with Parkinson’s disease and the disease group (ADrp) included people with AD-related mild cognitive impairment (referred to as MCIAD in the text) and with Alzheimer’s disease. The method was tested on four cohorts: (1) The unseen 1.5 T ADNI cohort (30% of the entire 1.5 T cohort, made up of 65 CN, 62 MCIAD, 54 AD, 28 FTD and 25 PD); (2) The unseen 1.5 T dataset: 64 people obtained from the Open Access Series of Imaging Studied (OASIS) consortium with baseline T1w MRI scan and the mini-mental state examination (MMSE) score (53 CN and 11 AD); (3) The unseen 3 T dataset: 402 people obtained from ADNI with T1w MRI scan, MMSE, logical memory delayed recall total (LDELTOTAL), Aβ, τ and pτ (172 CN, 161 MCIAD and 69 AD); (4) The ‘real-world’ memory clinic cohort (IMC cohort): 83 patients with atypical presentations who underwent clinical Amyloid PET imaging as part of their diagnostic workup with a 1.5 T T1w MRI scan (45 amyloid-negative (AMY−) and 38 amyloid-positive (AMY+)) and LDELTOTAL and MMSE scores (for a subgroup of 22 people: 11 AMY− and 11 AMY+).

Fig. 1

Overview of the study design and two-step least absolute shrinkage and selection operator (LASSO) approach.

Data used in this work were obtained from ADNI database, the OASIS consortium and the hospital memory clinic (IMC Cohort). Age-matched T1w MRI images were collected and segmented into 115 brain regions using the FreeSurfer’s recon-all function. Isotropic (1 × 1 × 1) T1w MRI scans and their brain masks were used for the radiomic analysis in a combined double step approach. After the selection and the standardisation of features, a first least absolute shrinkage and selection operator (LASSO1) was trained to classify people into those without and with AD-related pathology (nADrp and ADrp). Within the last group, a second LASSO (LASSO2) was trained to characterise patients with a mild cognitive impairment due to AD (MCIAD) from AD patients. The model was also integrated with cognitive scores (MMSE and LDELTOTAL) and CSF-based biomarkers (Aβ, τ and pτ). As the final algorithm was to be used to discriminate between ADrp and nADrp, combined healthy controls and patients affected by other non-AD pathologies (e.g. Frontotemporal dementia and Parkinson’s disease dementia) were combined into one group referred to as non-AD-related pathology group. Initial analysis of T2w MRI data did not yield discriminatory information, so only T1w MRI data is reported.

For the IMC cohort, we received ethical approval from the Camden and Kings Cross UK Research Ethics Committee (IRAS n. 273966) to perform retrospective anonymised and unlinked analysis of all clinical data (including MR images), provided that these were anonymised at source by a member of the clinical care team. In particular, the study protocol states: 'For all patients undergoing Amyloid PET at Imperial College Healthcare NHS Trust (ICHT) from December 2013 to January 2023 we will perform retrospective anonymised and unlinked analysis of clinically collected data. This will be anonymised at source by members of the clinical care team. The data will be unlinked and there will be no prospective element to this data collection.' Informed consent was waived, as is the case for retrospective analysis of anonymised imaging data.

Data for ADNI and OASIS are openly available upon registration of investigator interest. All participants provided informed consent. Details about the Ethics statement of the ADNI study population can be found at: https://adni.loni.usc.edu. Details about the Ethics statement of the OASIS study population can be found at: https://www.oasis-brains.org/#data. Protocols for data collection and the list of institutions who approved data collection can be found at https://adni.loni.usc.edu/methods/documents/ for ADNI. OASIS is made available by the Washington University Alzheimer’s Disease Research Center, the Howard Hughes Medical Institute (HHMI) at Harvard University, the Neuroinformatics Research Group (NRG) at Washington University School of Medicine, and the Biomedical Informatics Research Network (BIRN).

MRI segmentation and radiomic analysis

T1w MRI images were segmented to brain masks of 115 sub-regions using the FreeSurfer’s recon-all function (45 regions obtained from the segmentation of the white matter +70 subcortical regions obtained from the additional segmentation of the cortex)[25,26]. Before segmentation, this function performs many pre-processing steps, including bias correction, image sampling and coregistration; the steps and brain regions extracted are summarised in Supplementary Table 1. The multi-regional brain masks were post-processed for the extraction of 656 features for each region using in-house software (TexLAB 2.0), which runs on MATLAB[16]. The extracted features are related to the shape and size, intensity, texture and wavelet decompositions of isotropic (1 × 1 × 1) T1w MRI scans (Supplementary Data 2). The standardised radiomic features with a false discovery rate (FDR) <5% were selected as the input for the LASSO. Tenfold cross-validation was performed to select lambda which yielded the minimum cross-validated mean squared error. The weighted sum of the selected features gave the Alzheimer’s predictive Vector, ApV. For improving the model performance, the method was integrated with two cognitive measurements (MMSE and LDELTOTAL) and three CSF-based biomarkers (Aβ, τ and pτ). The result was a second predictive vector: ApVs.

The model is composed of two steps:

In the first stage of the classification, the algorithm works on the discrimination of people with an Alzheimer related pathology. The two inputs to the LASSO1 are the nADrp group, which includes healthy controls and people with Parkinson’s and frontotemporal dementia, and the ADrp group, which includes people with MCIAD and AD. The result of the LASSO is a reduced number of features/regions with their correspondent weights. The weighted sum of regions/features gives the ApV1 (ApV1s with the inclusion of cognitive scores and CSF related biomarkers). People classified as not- nADrp are used as inputs for the second stage of the classification.

In the second stage of the classification, the algorithm works on the distinction between people with an AD-related mild cognitive impairment and with Alzheimer’s disease. The LASSO2 performs a weighted sum of selected features/regions and gives the ApV2 (ApV2s with the inclusion of cognitive scores and CSF related biomarkers) which characterise a prodromal from a late phase of AD.

The performance of the algorithm was tested using two methods. In Method A, the features extracted from the 45-region brain mask (alone and together with cognitive/CSF scores) were used and, in Method B, features extracted from the (45 + 70)-region brain mask (alone and together with cognitive/CSF scores) were used. Based on the accuracy and the accuracy/AUC values, Method B was chosen for the computation of the ApV1, and Method A was chosen for the computation of ApV1s, ApV2 and ApV2s (Table 1).

Table 1

Methods comparison.

			AUC	Threshold	Specificity	Sensitivity	Accuracy	PPV	NPV
METHOD A (45 regions)	nADrp vs ADrp	T1w MRI	0.9047	−0.0387	0.8224	0.8362	0.8284	0.7823	0.8681
		T1w MRI + scores	0.9971	−0.1969	0.9671	0.9310	0.9554	0.9558	0.9484
	MCIAD vs AD	T1w MRI	0.7942	0.0648	1.0000	0.5185	0.7759	1.0000	0.7045
		T1w MRI + scores	0.9656	0.8184	0.9384	0.8583	0.8633	0.9237	0.8839
METHOD B(45 + 70 regions)	nADrp vs ADrp	T1w MRI	0.9920	0.0938	0.9831	0.9741	0.9786	0.9826	0.9748
		T1w MRI + scores	0.9859	0.6318	0.9830	0.9741	0.9786	0.9826	0.9747
	MCIAD vs AD	T1w MRI	0.7984	0.2554	0.9516	0.5556	0.7672	0.9091	0.7108
		T1w MRI + scores	0.9367	0.1428	0.8871	0.8333	0.8621	0.8654	0.8594

The classification between nADrp and ADrp, as well as the classification between MCIAD and AD patients were tested with two methods.

With Method A, the algorithm received as input features extracted from the 45 brain regions resulting from segmentation of the white matter (without and with the CSF/cognitive scores). Method B considered the features extracted from the 70 subcortical regions (without and with the CSF/cognitive scores).

Genomic analysis

Six genome-wide association study (GWAS) analyses were performed across three phenotypes (nADrp, MCIAD, AD) derived from three variables (original label (ADNI), ApV and ApVs). One GWAS was performed for nADrp vs MCIAD and another GWAS for nADrp vs AD across all five variables. APOE4 allele status was provided by ADNI APOE genotype dataset. All the GWAS analyses were adjusted for age and gender using the GWASTools R package (v1.36). Each GWAS analysis calculated the main effects of all single-nucleotide polymorphisms (SNPs) on the target label (MCIAD /AD). For all GWAS the empirical p values were based on the Wald statistic[27]. Manhattan plots were used to visualise GWAS results.

Statistics and reproducibility

Standard statistical analysis was applied to all the figures as appropriate and indicated in the figure legends. All samples were used once. Multiple testing was corrected with the FDR method. All the statistical analyses were conducted in Matlab R2019b.

Table 2

Diagnostic performance of the Alzheimer’s predictive vector ApV1 and ApV1s.

	Training 1.5 T ADNI dataset		Unseen 1.5 T ADNI dataset		Unseen 1.5 T OASIS dataset		Unseen 3 T ADNI dataset
	ApV1	ApV1s	ApV1	ApV1s	ApV1	ApV1s	ApV1	ApV1s	Volume of hippocampus	Aβ
AUC	0.9981	0.9971	0.9786	0.9490	0.6706	0.6801	0.6533	0.5192	0.7790	0.5045
Threshold	0.0938	−0.1969	0.0938	−0.1969	0.0938	−0.1969	0.0938	−0.1969	−0.1132	192
Specificity	0.9818	0.9669	0.9831	0.9671	0.8868	0.9057	0.9127	0.8081	0.2273	0.0091
Sensitivity	0.9819	0.9780	0.9741	0.9310	0.4545	0.4545	0.1739	0.2304	0.2941	1
Accuracy	0.9836	0.9728	0.9786	0.9554	0.8125	0.8281	0.4900	0.4776	0.2626	0.6236
NPV	0.9855	0.9750	0.9748	0.9484	0.8868	0.8889	0.4524	0.4398	0.2227	1
PPV	0.9818	0.9709	0.9826	0.9558	0.4545	0.500	0.7272	0.6162	0.2996	0.6223
LR +	54.4364	29.5952	57.4741	28.3034	4.0151	4.8182	1.9942	1.2010	0.3806	1.0009
LR−	0.0149	0.0868	0.0263	0.0713	0.6151	0.6023	0.9050	0.9522	3.1059	0
Yi	0.9672	0.9450	0.9572	0.8981	0.3413	0.3602	0.0867	0.0385	−0.4786	0.0092
DOR	3653.4	1301.6	2184.6	396.9	6.5278	8	2.2035	1.2612	0.1225	NA

Diagnostic performance of ApV1 and ApV1s was evaluated in the 1.5 T training dataset (ADNI), the unseen 1.5 T ADNI, 1.5 T OASIS and 3 T ADNI datasets. The performance of the ApV is also compared to the current clinically used measure of hippocampal volume in the discrimination between nADrp and ADrp patients, and CSF Aβ in the discrimination between CN and ADrp.

In the testing test, AUC values were generated from sensitivity and specificity[62].

DOR diagnostic odds ratio, Yi Youden index value, LR+ positive likelihood ratio, LR− negative likelihood ratio, NA undefined values derived from the division by zero, NPV negative predictive value, PPV positive predictive value.

Table 3

Diagnostic performance of the Alzheimer’s predictive vectors ApV2 and ApV2s.

	Training 1.5 T ADNI dataset		Unseen 1.5 T ADNI dataset		Unseen 3 T ADNI dataset
	ApV2	ApV2s	ApV2	ApV2s	ApV2	ApV2s	Volume of hippocampus	Aβ
AUC	0.8580	0.9656	0.7258	0.8983	0.5072	0.7111	0.5345	0.5
Threshold	0.3017	0.8184	0.3017	0.8184	0.3017	0.8184	−0.7827	192
Specificity	0.9863	0.9384	0.9516	0.9384	1	0.9875	0.3387	0
Sensitivity	0.5590	0.8583	0.5000	0.8583	0.0289	0.4347	0.7593	1
Accuracy	0.7875	0.9011	0.7863	0.8633	0.6296	0.8217	0.5345	0.4887
NPV	0.7200	0.8839	0.6860	0.8839	0.7061	0.8030	0.6176	NA
PPV	0.9726	0.9237	0.9000	0.9237	1	0.9375	0.5000	0.4887
LR +	40.8110	13.9230	10.3333	13.9230	NA	35.0000	1.1481	1
LR−	0.4471	0.1510	0.5254	0.1510	0.9710	0.5723	0.7108	NA
Yi	0.5454	0.7966	0.4516	0.7966	0.0289	0.4223	0.0980	0
DOR	91.2857	92.1790	19.6667	92.1790	NA	61.1538	1.6154	NA

Diagnostic performance of ApV2 and ApV2s evaluated in the 1.5 T training dataset (ADNI), the unseen 1.5 T ADNI and 3 T ADNI datasets compared to the volume of the hippocampus and Aβ in the discrimination between MCIAD and AD patients. *Of note, the measurements of diagnostic accuracy of Aβ are obtained with the application of the established cut-off values (Shaw et al.).

In the testing test, AUC values were generated from sensitivity and specificity[62].

DOR diagnostic odds ratio, Yi Youden index value, LR+ positive likelihood ratio, LR− negative likelihood ratio, NA undefined values derived from the division by zero, NPV negative predictive value, PPV positive predictive.

Table 4

Test on the diagnostic performance of the algorithm.

		AUC	Threshold	Specificity	Sensitivity	Accuracy	PPV	NPV
nADrp vs ADrp	train	0.9981	0.0938	0.9819	0.9853	0.9836	0.9818	0.9855
	test	0.9920	0.0938	0.9831	0.9741	0.9786	0.9826	0.9748
CN vs ADrp	train	1.0000	−0.1109	0.9934	1.0000	0.9976	0.9964	1.0000
	test	1.0000	−0.1109	1.0000	0.9828	0.9890	1.0000	0.9701
CN vs MCIAD	train	1.0000	0.0722	1.0000	0.9932	0.9966	1.0000	0.9934
	test	1.0000	0.0722	1.0000	0.9839	0.9921	1.0000	0.9848
CN vs AD	train	0.9999	−0.1109	0.9934	1.0000	0.9964	0.9922	1.0000
	test	1.0000	−0.1109	1.0000	0.9815	0.9916	1.0000	0.9848

The two inputs to the LASSO1 are the nADrp group, which includes healthy controls and people with Parkinson’s and frontotemporal disease, and the ADrp group, which includes people with MCIAD and AD. The diagnostic performance of the algorithm was tested when the classification is computed between the ADrp group and healthy people, between CN and MCIAD and CN and AD patients.