Endocytoscopyによる超拡大内視鏡診断

Endocytoscopic Diagnosis of Colorectal Neoplasms

要旨●超拡大内視鏡（EC）は，生体内で腺腔構造および核形態を直接観察できる唯一の内視鏡モダリティであり，大腸腫瘍診断における“optical biopsy”を現実のものとした．従来のpit pattern診断や狭帯域光観察（NBI）拡大観察は，腺管構造や表層微細血管構築といった構造異型の評価に優れる一方で，細胞異型そのものを直接評価することは困難であった．ECは最大520倍の超拡大観察により，核腫大，核配列，腺腔不明瞭化などの細胞レベルの異型をリアルタイムに描出し，腫瘍性診断および深達度診断の精度向上に寄与する．本稿では，大腸腫瘍診断におけるECの技術的背景と診断体系を概説するとともに，pit pattern診断との相補関係，診断限界，さらに近年臨床導入が進むAI診断支援（EndoBRAIN®シリーズ）の位置付けと今後の展望について述べる．

　Endocytoscopy（EC）is an ultra-high-magnification endoscopic modality that enables real-time cellular observation of the colorectal mucosa. Direct visualization of the glandular architecture and nuclear morphology in EC provides information beyond conventional magnifying endoscopy. Consequently, it has been proposed as a form of “optical biopsy.”

　Pit pattern diagnosis, established via chromoendoscopy, remains a cornerstone method for the qualitative diagnosis and invasion depth estimation of colorectal neoplasms. However, lesions with type V or especially type VI pit patterns are often a diagnostic gray zone, where accurate assessment of submucosal deep invasion remains challenging.

　Recent studies have shown that EC does not considerably improve diagnostic differentiation between neoplastic and nonneoplastic lesions beyond pit pattern diagnosis. Nevertheless, EC markedly improves the specificity and overall accuracy in diagnosing deeply invasive submucosal carcinoma, suggesting a complementary contribution to preventing overdiagnosis. EC3b findings, such as loss of glandular lumen and marked nuclear atypia, may reflect focal high-grade malignancy, which cannot be fully appraised by surface structural assessment.

　In this review, we summarize the technical background of EC, its diagnostic classification, and its contributions to pit pattern diagnosis. We also discuss the clinical significance and limitations of EC by analyzing representative cases and published evidence and address the potential advantages of artificial intelligence-assisted diagnosis. Ultimately, the effective integration of EC into existing diagnostic algorithms based on a clear understanding of its strengths and limitations is essential for optimizing clinical decision-making in colorectal tumor management.