Neuro spectra: Multi-domain attention-guided feature calibration for Parkinson’s disease prediction

PD (Parkinson’s disease) is considered the advanced neurodegenerative disorder that affects the movements as the dopamine-producing neurons in the substantia nigra location present in the brain get reduced. Since the accurate basis of PD is the mixture of genetic and environmental factors, PD results in the motor-related symptoms. However, the earlier detection is significant for the management of indications and slow growth and the diagnosis is difficult because it is identical to several neurodegenerative situations. The conventional techniques include the clinical assessments, patients’ profiles and the MRI, SPECT and PET scans. Since the MRI’s focus on the structural differences restricts the efficiency in the earlier phase of PD detection. Therefore, the ML and DL models are utilized in the classification of MRI scan images, as they face issues in the computations of features through several medical images and classification. Despite the enhancements, several difficulties exist such as restricted, labeled and noisy datasets. Hence, the study proposes a Modified EfficientNetV2-CMAFM (Cross Modal Attention Fusion Module): Dual-Domain Attentive Residual Network for the detection of PD in MRI scans with the utilization of the PPMI and NTUA Parkinson’s disease datasets. In contrast to the existing models, the proposed Modified EfficientNetV2-CMAFM: Dual-Domain Attentive Residual Network achieves superior performance and early detection. The proposed Modified EfficientNetV2-CMAFM: Dual-Domain Attentive Residual Network is evaluated using the performance metrics such as accuracy, precision, recall and F1 score, in which the proposed model obtains higher metrics values.