The research conducted at Trita Hospital in Tehran, Iran involved collecting data from various imaging modalities to diagnose prostate cancer. The dataset included T1, T2-tra, T2-sag, T2-cor, DW images, and ADC maps. Biopsy results and clinical information such as age, antigen levels, and prostate dimensions were also collected. Patients were labeled with PI-RADS-V2 scores and pathological data based on Gleason scoring. Data preprocessing involved normalization and augmentation techniques to increase the dataset size for training CNN models. Four CNN models were trained with transfer learning using ResNet50 architecture. Features extracted from patient images were combined with clinical data for further analysis using neural networks. The final model architecture integrated clinical, pathological, and imaging data for accurate cancer diagnosis. Figures displayed the overall model and the ANN network architecture. The study demonstrated the importance of combining imaging data with clinical information for improved prostate cancer diagnosis.
Source link
Source link: https://www.nature.com/articles/s41598-024-65354-0
Deep learning for prostate cancer diagnosis using multi-parametric MRI #CancerDiagnosis
