Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders
Magnetic resonance imaging (MRI) nowadays plays an important role in the identification of brain underpinnings in a wide range of neuropsychiatric disorders, including Autism Spectrum Disorders (ASD). Characterizing the hallmarks in these pathologies is not a straightforward task and machine learnin...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/13/11/6486 |
| _version_ | 1797597988392009728 |
|---|---|
| author | Federico Campo Alessandra Retico Sara Calderoni Piernicola Oliva |
| author_facet | Federico Campo Alessandra Retico Sara Calderoni Piernicola Oliva |
| author_sort | Federico Campo |
| collection | DOAJ |
| description | Magnetic resonance imaging (MRI) nowadays plays an important role in the identification of brain underpinnings in a wide range of neuropsychiatric disorders, including Autism Spectrum Disorders (ASD). Characterizing the hallmarks in these pathologies is not a straightforward task and machine learning (ML) is certainly one of the most promising tools for addressing complex and non-linear problems. ML algorithms and, in particular, deep neural networks (DNNs), need large datasets in order to be properly trained and thus ensure generalization capabilities on new data. Large datasets can be obtained by collecting images from different centers, thus bringing unavoidable biases in the analysis due to differences in hardware and scanning protocols between different centers. In this work, we dealt with the issue of multicenter MRI data harmonization by comparing two different approaches: the analytical ComBat-GAM procedure, whose effectiveness is already documented in the literature, and an originally developed site-adversarial deep neural network (ad-DNN). The latter aims to perform a classification task while simultaneously searching for site-relevant patterns in order to make predictions free from site-related biases. As a case study, we implemented DNN and ad-DNN classifiers to distinguish subjects with ASD with respect to typical developing controls based on functional connectivity measures derived from data of the multicenter ABIDE collection. The classification performance of the proposed ad-DNN, measured in terms of the area under the ROC curve (AUC), achieved the value of AUC = <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.70</mn><mo>±</mo><mn>0.03</mn></mrow></semantics></math></inline-formula>, which is comparable to that obtained by a DNN on data harmonized according to the analytical procedure (AUC = <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.71</mn><mo>±</mo><mn>0.01</mn></mrow></semantics></math></inline-formula>). The relevant functional connectivity alterations identified by both procedures showed an agreement between each other and with the patterns of neuroanatomical alterations previously detected in the same cohort of subjects. |
| first_indexed | 2024-03-11T03:12:04Z |
| format | Article |
| id | doaj.art-f9fbd272fa774e7ba9d3576a8b2dc1d4 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-11T03:12:04Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-f9fbd272fa774e7ba9d3576a8b2dc1d42023-11-18T07:32:47ZengMDPI AGApplied Sciences2076-34172023-05-011311648610.3390/app13116486Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum DisordersFederico Campo0Alessandra Retico1Sara Calderoni2Piernicola Oliva3Department of Physics, University of Pisa, 56127 Pisa, ItalyNational Institute for Nuclear Physics (INFN), Pisa Division, 56127 Pisa, ItalyDevelopmental Psychiatry Unit, IRCCS Stella Maris Foundation, 56127 Pisa, ItalyDepartment of Chemical, Physical, Mathematical and Natural Sciences, University of Sassari, 07100 Sassari, ItalyMagnetic resonance imaging (MRI) nowadays plays an important role in the identification of brain underpinnings in a wide range of neuropsychiatric disorders, including Autism Spectrum Disorders (ASD). Characterizing the hallmarks in these pathologies is not a straightforward task and machine learning (ML) is certainly one of the most promising tools for addressing complex and non-linear problems. ML algorithms and, in particular, deep neural networks (DNNs), need large datasets in order to be properly trained and thus ensure generalization capabilities on new data. Large datasets can be obtained by collecting images from different centers, thus bringing unavoidable biases in the analysis due to differences in hardware and scanning protocols between different centers. In this work, we dealt with the issue of multicenter MRI data harmonization by comparing two different approaches: the analytical ComBat-GAM procedure, whose effectiveness is already documented in the literature, and an originally developed site-adversarial deep neural network (ad-DNN). The latter aims to perform a classification task while simultaneously searching for site-relevant patterns in order to make predictions free from site-related biases. As a case study, we implemented DNN and ad-DNN classifiers to distinguish subjects with ASD with respect to typical developing controls based on functional connectivity measures derived from data of the multicenter ABIDE collection. The classification performance of the proposed ad-DNN, measured in terms of the area under the ROC curve (AUC), achieved the value of AUC = <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.70</mn><mo>±</mo><mn>0.03</mn></mrow></semantics></math></inline-formula>, which is comparable to that obtained by a DNN on data harmonized according to the analytical procedure (AUC = <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.71</mn><mo>±</mo><mn>0.01</mn></mrow></semantics></math></inline-formula>). The relevant functional connectivity alterations identified by both procedures showed an agreement between each other and with the patterns of neuroanatomical alterations previously detected in the same cohort of subjects.https://www.mdpi.com/2076-3417/13/11/6486brain connectivitymachine learningadversarial learningautism spectrum disorders (ASD)multi-site harmonizationexplainable AI (XAI) |
| spellingShingle | Federico Campo Alessandra Retico Sara Calderoni Piernicola Oliva Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders Applied Sciences brain connectivity machine learning adversarial learning autism spectrum disorders (ASD) multi-site harmonization explainable AI (XAI) |
| title | Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders |
| title_full | Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders |
| title_fullStr | Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders |
| title_full_unstemmed | Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders |
| title_short | Multi-Site MRI Data Harmonization with an Adversarial Learning Approach: Implementation to the Study of Brain Connectivity in Autism Spectrum Disorders |
| title_sort | multi site mri data harmonization with an adversarial learning approach implementation to the study of brain connectivity in autism spectrum disorders |
| topic | brain connectivity machine learning adversarial learning autism spectrum disorders (ASD) multi-site harmonization explainable AI (XAI) |
| url | https://www.mdpi.com/2076-3417/13/11/6486 |
| work_keys_str_mv | AT federicocampo multisitemridataharmonizationwithanadversariallearningapproachimplementationtothestudyofbrainconnectivityinautismspectrumdisorders AT alessandraretico multisitemridataharmonizationwithanadversariallearningapproachimplementationtothestudyofbrainconnectivityinautismspectrumdisorders AT saracalderoni multisitemridataharmonizationwithanadversariallearningapproachimplementationtothestudyofbrainconnectivityinautismspectrumdisorders AT piernicolaoliva multisitemridataharmonizationwithanadversariallearningapproachimplementationtothestudyofbrainconnectivityinautismspectrumdisorders |