FPGA實現醫學影像即時辨識嵌入式系統 = FPGA-based embedded system for real-time medical image recognition / 張耀中.
- 作者： 張耀中
- 其他題名：
- FPGA-based embedded system for real-time medical image recognition
- 主題： 第四代低功率雙倍資料傳輸速率同步動態隨機存取記憶體控制器 安全數位記憶卡控制器 現場可程式化邏輯閘陣列. , LPDDR4 Controller SD-Card Controller FPGA.
- URL：
電子資源
- 一般註：指導教授: 陳瓊安. 學年度: 113.
- 書目註：參考書目： 葉.
-
讀者標籤：
- 系統號： 005185164 | 機讀編目格式
館藏資訊
摘要註
黑色素瘤在診斷過程中，皮膚科醫師通常會透過觀察黑痣的外觀特徵，包括形狀不對稱、邊緣不規則、顏色不均勻、病灶大於 6mm 及分布位置，初步判斷其為高機率黑色素瘤的可能性。在醫師建議切除病灶位置前，醫師會以相機拍攝黑痣外觀，作為資料集收集與後續分析與化驗之用。為協助醫學影像判讀，本研究於 FPGA 嵌入式系統中建立一套醫學影像辨識與無線傳輸系統架構，使黑色素瘤影像得以使用MicroSD Card 傳輸至 FPGA 板上進行黑色素瘤影像分析。最終本文整合 SD-Card Controller、LPDDR4 Controller 與 Image Data Buffer 數位電路，電路硬體資源使用量 Flip Flops 為 5,539、Look Up Tables 為 2,631， ASIC Gate Count 為 15.71 k。色彩影像大小 512 × 512 pixel × 3 channel 於本文影像資料傳輸系統之傳輸時間為 378ms。. During the diagnostic process for melanoma, dermatologists typically evaluate the visible characteristics of moles—such as asymmetry in shape, irregular borders, uneven pigmentation, lesion size exceeding 6 mm, and their distribution on the body—to initially assess the likelihood of malignancy. Prior to recommending surgical excision, the physician captures images of the suspected lesion using a camera, which are then used for dataset collection, further analysis, and pathological examination. To support medical image interpretation, this study implements a medical image recognition and wireless transmission system on an FPGA-based embedded platform. Melanoma images stored on a MicroSD card are transmitted to the FPGA board for real-time image analysis. The system incorporates three core digital circuit modules: an SD-Card Controller, an LPDDR4 Controller, and an Image Data Buffer. Hardware resource utilization includes 5,539 flip- flops, 2,631 look-up tables, and a total ASIC gate count of 15.71k. For a color image sized 512 × 512 pixels across three channels, the proposed transmission system achieves a data transfer time of 378 ms..