Department of Gynaecology and Obstetrics, Emory University School of Medicine, Atlanta, GA, USA.
Department of Mathematics, Rensselaer Polytechnic Institute, Troy, NY, USA.
J Med Eng Technol. 2021 Nov;45(8):606-613. doi: 10.1080/03091902.2021.1936674. Epub 2021 Jul 6.
This study hypothesised that benign and tumour-bearing uterine tissue could be differentiated by their unique electrical bioimpedance patterns, with the aid of artificial intelligence. Twenty whole, uterine specimens were obtained at the time of hysterectomy. A total of 11 benign and 9 malignant specimens were studied. A uterine bioimpedance probe was designed to measure the tissue between the endometrial and serosal layers of the uterus. The impedance data was then analysed with multiple instance learning and principal component analysis, forms of artificial intelligence. Final pathology results for the specimens included uterine sarcoma, adenocarcinoma, carcinosarcoma, and high-grade serous carcinoma. The analysis correctly identified 78% (7/9) of the malignant specimens and 82% (9/11) of the benign specimens. The overall accuracy of our analysis was 80%. Our results demonstrate distinction between electrical impedance properties of malignant and benign uterine specimens. Bioimpedance and artificial intelligence may have potential implications in risk assessment of patients and may subsequently guide surgical decision-making regarding route of organ removal.
本研究假设，借助人工智能，良性和肿瘤性子宫组织可以通过其独特的电生物阻抗模式进行区分。在子宫切除术时获得了 20 个完整的子宫标本。共研究了 11 个良性和 9 个恶性标本。设计了一种子宫生物阻抗探头来测量子宫的子宫内膜层和浆膜层之间的组织。然后使用人工智能的多实例学习和主成分分析形式对阻抗数据进行分析。标本的最终病理结果包括子宫肉瘤、腺癌、癌肉瘤和高级别浆液性癌。分析正确识别了 78%（9/11）的恶性标本和 82%（7/9）的良性标本。我们分析的总准确率为 80%。我们的研究结果表明，恶性和良性子宫标本的电阻抗特性存在差异。生物阻抗和人工智能可能对患者的风险评估有潜在影响，并可能随后指导手术切除器官的途径。
