百洑傳息BIOFLUMEN & XENSOR LAB ← 線上工具 / Online ToolsNYCU · MEDTECH VIRTUAL LAB

MICROBIOLOGY · MICROSCOPY · EDUCATIONAL v0.1.1

Web-Gram Stain Lite

革蘭氏染色影像分析 Gram Stain Microscopy Teaching Analyzer

先觀察染色反應與菌形，提交你的判讀；再以瀏覽器內的色彩分割及形態特徵分析顯微鏡照片，揭示系統建議，並逐一校正候選標記。適用醫技系臨床微生物學與鏡檢課的操作示範。Interpret Gram reaction and morphology before revealing rule-based suggestions. This browser-only teaching tool detects stained regions, measures simple shape features, and supports manual marker review and export.

教學研究原型 · Educational prototype本版使用傳統影像處理規則，沒有載入 AI 模型，也尚未完成真實臨床影像的敏感度、特異度或計數驗證。染色沉澱、宿主細胞、菌體群聚與焦距可能產生偽陽性或漏判。系統建議須由教師或檢驗人員複核，勿作診斷、菌種鑑定或用藥依據。
Rule-based image segmentation; no deployed AI model or clinical accuracy validation. Stain precipitate, host cells, overlapping organisms and focus can distort results. Educational and research use only; never use for clinical diagnosis or treatment decisions.

01 載入染色影像 Upload & Demo

建議使用油鏡清晰視野、白色或淡色背景、光線均勻的照片。輸入影像在瀏覽器內縮放至最多 1,100 × 770 px；細小菌體可能在縮放後失去影像特徵。原始檔不會傳至本站伺服器。Use a sharp, evenly lit oil-immersion image with a pale background. Processing is local; images are resized to fit within 1,100 × 770 px, and tiny organisms may lose detail.

02 學生先行判讀 Student Interpretation

先選擇你觀察到的主要染色反應與菌形；提交後鎖定答案，待分析完成才顯示系統建議。Record your interpretation first. Your answer is locked when submitted, enabling a side-by-side comparison after analysis.

尚未提交學生判讀 / No student response yet.

03 分色與候選分析 Rule-based Detection

04 標註覆核 Manual Annotation

ROI 模式在圖上拖曳框選分析視野；提交學生判讀並揭示結果後，選擇類別點擊標記可改類型，點空白可新增標記，刪除模式可移除標記。更換 ROI 會清除原有分析及作答。Drag a rectangle to select an ROI. After revealing results, tap to relabel or add a marker; Erase removes it. Changing the ROI resets the analysis and student response.

請上傳染色影像或載入合成範例。Upload a Gram stain image or load a synthetic example.

05 顯微影像 Image Workspace

G+ 球菌G− 球菌G+ 桿菌G− 桿菌待確認

色圈表示程式篩出的候選連通色塊，大小與位置僅供人工覆核。排列方式、菌種、G-variable 與 yeast 尚無可靠自動分類。Colored outlines represent candidate connected components only. Arrangement, species, Gram-variable reactions, and yeasts are not automatically classified.

判讀原理 Method

本版根據 RGB 色差尋找紫色及粉紅色色塊，再用連通區面積及二階矩（major/minor axis ratio）粗略區分球形與桿形。形狀臨界區與過大候選會標示待確認；結果隨門檻、背景及染色而變動。RGB chromatic contrast detects purple/pink pixels. Connected components and second-moment axis ratios offer provisional cocci/rod suggestions. Ambiguous shapes are flagged for review.

臨床限制 Scope & Limitations

這個簡化模型對真實抹片的染色沉澱、厚薄差、背景細胞與密集群聚缺乏驗證。學生仍需辨識菌體大小、排列、染色品質及檢體背景。未包含病原菌種、Gram-variable、真菌、混合菌群的可靠推論，亦未提供染色品質分數。The detector lacks validation on clinical smears. Precipitates, dense growth, cell debris, and variations in stain quality can confound predictions. Manual microscope review remains essential.

相關文獻與公開資料 Research & Open Data

Wang X et al. (2024). A Clinical Bacterial Dataset for Deep Learning in Microbiological Rapid On-Site Evaluation. Scientific Data 11:608. DOI · Public dataset.

An annotated dataset of Gram stains from positive blood cultures. Scientific Data (2026). DOI · CC BY 4.0 dataset.

這些文獻提供後續模型訓練及教學影像標註的研究參考。本站 V0.1 採原創 JavaScript 程式與程式生成的合成示範圖，未引入上述論文的影像、模型權重或第三方演算法原始碼。The references inform future validation and model research. This original JavaScript prototype includes no externally sourced microscopy image, pretrained model, or copied third-party algorithm.