Abstract：
Basal cell carcinoma (BCC), the most common keratinocyte cancer, presents a substantial public health challenge due to its high prevalence. Traditional diagnostic methods, which rely on visual examination and histopathological analysis, do not include metabolomic data. This exploratory study aims to molecularly characterize BCC and diagnose tumour tissue by applying matrix-assisted laser desorption ionization mass spectrometry imaging (MALDI-MSI) and machine learning (ML). BCC tumour development was induced in a mouse model and tissue sections containing BCC (n = 12) were analysed. The study design involved three phases: (i) Model training, (ii) Model validation and (iii) Metabolomic analysis. The ML algorithm was trained on MS data extracted and labelled in accordance with histopathology. An overall classification accuracy of 99.0% was reached for the labelled data. Classification of unlabelled tissue areas aligned with the evaluation of a certified Mohs surgeon for 99.9% of the total tissue area, underscoring the model\'s high sensitivity and specificity in identifying BCC. Tentative metabolite identifications were assigned to 189 signals of importance for the recognition of BCC, each indicating a potential tumour marker of diagnostic value. These findings demonstrate the potential for MALDI-MSI coupled with ML to characterize the metabolomic profile of BCC and to diagnose tumour tissue with high sensitivity and specificity. Further studies are needed to explore the potential of implementing integrated MS and automated analyses in the clinical setting.
摘要：
基底细胞癌（BCC），最常见的角质形成细胞癌，由于其高患病率，提出了重大的公共卫生挑战。传统的诊断方法，依靠视觉检查和组织病理学分析，不包括代谢组学数据。这项探索性研究旨在通过应用基质辅助激光解吸电离质谱成像（MALDI-MSI）和机器学习（ML）对BCC进行分子表征和诊断肿瘤组织。在小鼠模型中诱导BCC肿瘤发展,并分析含有BCC的组织切片(n=12)。研究设计包括三个阶段：(I)模型训练，(ii)模型验证和(iii)代谢组学分析。在根据组织病理学提取和标记的MS数据上训练ML算法。标记数据的总体分类准确度达到99.0%。未标记组织区域的分类与经过认证的Mohs外科医生的评估一致，占总组织面积的99.9％，强调了该模型在识别BCC方面的高灵敏度和特异性。初步代谢物鉴定被分配给189个重要信号，用于识别BCC，每个都表明潜在的具有诊断价值的肿瘤标志物。这些发现证明了MALDI-MSI与ML结合表征BCC的代谢组学特征并以高灵敏度和特异性诊断肿瘤组织的潜力。需要进一步的研究来探索在临床环境中实施集成MS和自动分析的潜力。
