Design of Lightweight Custom IP Core on FPGA to Discriminate Brain Tumors of MR Data

• Published in: Devices for Integrated Circuit pp. 494 - 499
• Main Authors: Chatterjee, Soumita, Pandit, Soumya, Das, Arpita
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.04.2025
• Subjects: Accuracy; Brain Tumor Detection; Brain tumors; Complexity theory; Custom IP Core; Feature extraction; Field programmable gate arrays; IP networks; Morphological Features; Power demand; PYNQ Z2 FPGA; Real-time systems; Support vector machines; SVM Classifier; Timing; Xilinx Vivado
• ISSN: 2996-3044
• DOI: 10.1109/DevIC63749.2025.11012554

In response to growing demand for efficient medical image processing methods on FPGA, this work presents the design of a custom IP core for brain tumor classification using morphological features. This IP core uses the ZYNQ 7000 Processing System on PYNQ Z2 platform. The present work focuses on minimization of classifier's t raining complexity by extracting 17 significant morphological features from tumor data. The tumors are then categorised with the best accuracy of 90% using a linear Support Vector Machine (SVM). As this streamlined methodology employs hand-crafted features to simplify the classification model, the proposed IP core uses only 1.92% of FPGA resources and 21 mW of power. The current work presents a potential design for scalable and hardware-accelerated medical imaging systems attaining minimal power with high efficiency.