Clear cell renal cell carcinoma (ccRCC) and chromophobe renal cell carcinoma (chRCC) are relatively common tumors that can have significant risk for mortality. Treatment and prognostication in renal cell carcinoma (RCC) are dependent upon correct histologic typing. ccRCC and chRCC are generally straightforward to diagnose based on histomorphology alone. However, high-grade ccRCC and chRCC can sometimes resemble each other morphologically, particularly in small biopsies. Multiple immunostains and/or colloidal iron stain are sometimes required to differentiate the two. Imaging mass spectrometry (IMS) allows simultaneous spatial mapping of thousands of biomarkers, using formalin-fixed paraffin-embedded tissue sections. In this study, we evaluate the ability of IMS to differentiate between World Health Organization/International Society for Urological Pathology grade 3 ccRCC and chRCC. IMS spectra from a training set of 14 ccRCC and 13 chRCC were evaluated via support vector machine algorithm with a linear kernel for machine learning, building a classification model. The classification model was applied to a separate validation set of 6 ccRCC and 6 chRCC, with 19 to 20, 150-μm diameter tumor foci in each case sampled by IMS. Most evaluated tumor foci were classified correctly as ccRCC versus chRCC (99% accuracy, kappa=0.98), demonstrating that IMS is an accurate tool in differentiating high-grade ccRCC and chRCC.

透明细胞肾细胞癌（ccRCC）和发色肾细胞癌（chRCC）是相对常见的肿瘤，可能具有重大的死亡风险。肾细胞癌（RCC）的治疗和预后取决于正确的组织学分型。ccRCC和chRCC通常仅凭组织形态即可直接诊断。但是，高级ccRCC和chRCC有时在形态上可能彼此相似，尤其是在小活检中。有时需要多种免疫染色和/或胶体铁染色来区分两者。成像质谱（IMS）允许使用福尔马林固定石蜡包埋的组织切片同时对数千个生物标记物进行空间定位。在这个研究中，我们评估了IMS区分世界卫生组织/国际泌尿外科病理学会3级ccRCC和chRCC的能力。通过带有线性核的机器学习的支持向量机算法，评估了来自14 ccRCC和13 chRCC训练集的IMS光谱，建立了分类模型。将分类模型应用于6 ccRCC和6 chRCC的单独验证集，每种情况下均由IMS采集19至20个直径为150μm的肿瘤灶。大多数评估的肿瘤灶被正确分类为ccRCC与chRCC（99％准确度，k = 0.98），表明IMS是区分高级ccRCC和chRCC的准确工具。通过带有线性核的机器学习的支持向量机算法，评估了来自14 ccRCC和13 chRCC训练集的IMS光谱，建立了分类模型。将分类模型应用于6 ccRCC和6 chRCC的单独验证集，每种情况下均由IMS采集19至20个直径为150μm的肿瘤灶。大多数评估的肿瘤灶被正确分类为ccRCC与chRCC（99％准确度，k = 0.98），表明IMS是区分高级ccRCC和chRCC的准确工具。通过带有线性核的机器学习的支持向量机算法，评估了来自14 ccRCC和13 chRCC训练集的IMS光谱，建立了分类模型。将分类模型应用于6 ccRCC和6 chRCC的单独验证集，每种情况下均由IMS采集19至20个直径为150μm的肿瘤灶。大多数评估的肿瘤灶被正确分类为ccRCC与chRCC（99％准确度，k = 0.98），表明IMS是区分高级ccRCC和chRCC的准确工具。