Parkinson Hastalığı Modellerinin Entegre Dopaminerjik Nöron Segmentasyonu ve Fenotipik Niceleme Yoluyla Otomatik Histolojik Değerlendirmesi

Araştırma özeti

Aşağıdaki metin kaynak makalenin lisansı doğrulanmış orijinal özetidir. cc by

Abstract Accurate quantification of dopaminergic neurons and their phenotypic state is essential for understanding neurodegeneration and evaluating therapeutic efficacy in Parkinson’s disease models. Traditional approaches, such as stereological counting, are labor-intensive and limited to structural assessment, while most automated methods neglect key biochemical markers like Tyrosine Hydroxylase (TH), a critical indicator of dopaminergic identity and viability. To overcome the limitations of existing methods that either focus solely on morphological segmentation or lack integrated biochemical analysis, a novel dual-head transformer-based architecture, SegDAFormer, is introduced in this research. It is specifically designed to jointly perform accurate dopaminergic neuron segmentation and fine-grained, pixel-wise quantification of TH expression from high-resolution histological images, thereby unifying structural and phenotypic assessment within a single, biologically interpretable framework. The model integrates a Hierarchical Transformer Encoder with a Multi-Scale Integrative Module, enabling effective fusion of local and global contextual features. Evaluated on a real TH-stained dataset, SegDAFormer achieves a Dice score of 93.12 ± 1.08%, counting error of 5.82 ± 1.45%, and TH intensity prediction error of MAE 7.96 ± 1.22, significantly outperforming existing baselines. Moreover, visual overlays confirm that predicted intensity maps reflect biologically meaningful gradients in neuronal phenotype. By bridging structural and biochemical analysis in a unified, interpretable framework, SegDAFormer advances automated histological assessment and holds strong potential for scalable applications in neurodegenerative research, drug screening, and digital pathology.

İlgili konular