Dual-Stream Deep Feature and Cell Phenotype Fusion Model for the Diagnosis of Myeloproliferative Neoplasms

Diagnosing myeloproliferative neoplasms (MPNs) is challenging because of the nuanced and overlapping clinical manifestations of the various subtypes. Precise classification is essential for effective treatment and management of these disorders. This study introduces a novel Dual-Stream Deep Feature and Cell Phenotype Fusion Model (DS-DFCPF) to improve the diagnosis of MPNs. The model integrates deep learning features from whole-slide images with detailed phenotypic data extracted from cellular components, particularly megakaryocytes, which are pivotal in MPN pathology. The DS-DFCPF uses a bifurcated approach, wherein one stream processes deep features from segmented whole-slide images using convolutional neural networks, and the other analyzes cell phenotype characteristics using advanced image processing techniques. The outputs of both streams are fused, significantly enhancing the model's capacity to discriminate between MPN subtypes. The efficacy of this model was rigorously evaluated through a series of experiments using a data set comprising 411 patient samples annotated with detailed clinical and histopathological information. Our results reveal that the DS-DFCPF significantly outperforms previous diagnostic models, offering a reliable and reproducible tool for MPN subtype differentiation. This model offers a promising new tool for pathologists and clinicians, providing a more accurate, efficient, and automated approach to diagnosing MPNs, thereby facilitating timely and tailored therapeutic interventions. This study not only underscores the potential of integrating multiple data streams in medical diagnostics but also establishes a benchmark for future innovations in the field of computational pathology.