探讨具有形态学人工智能(AI)识别图像的粪便分析仪在粪便肠道寄生虫筛查中的应用价值。

采用回顾性分析方法，收集2023年11月至2025年6月天门市第一人民医院检验科接收的、来自天门市各乡镇健康体检人群的粪便样本共9 845份，分别用常规人工显微镜方法与AI识别的粪便分析仪法对寄生虫的检出情况进行比较，并对结果的一致性采用Kappa检验。同时，分析两种方法对不同性别、寄生虫及乡镇人群的检出率。

AI识别共检出寄生虫阳性样本294例，阳性率2.99%(294/9 845)，人工复核确认阳性80例，阳性率0.81%，AI识别假阳性率2.19%(214/9 765)。常规人工显微镜检测阳性率0.56%(55/9 845)，漏检率31.25%(25/80)，两种方法检出率比较差异有统计学意义(χ²＝21.330，P<0.01)。AI识别＋人工复核与人工镜检在男女寄生虫检出率比较，均无统计学意义(χ²＝0.109，0.110；均P>0.05)。单纯AI识别与AI识别＋人工复核后寄生虫检出率达到中等一致率(Kappa＝0.417，P<0.01)，其敏感度高达98.75%，特异度97.80%，阳性预测值26.87%，阴性预测值99.99%；单纯人工显微镜检测与AI识别＋人工复核后寄生虫检出率高度一致(Kappa＝0.800，P<0.01)，其敏感度67.50%，特异度99.99%，阳性预测值98.18%，阴性预测值99.73%。两种方法均能检出华支睾吸虫卵、蓝氏贾第鞭毛虫、蛲虫卵、人芽囊原虫，AI识别还能检出粪类圆线虫、鞭虫卵、钩虫卵、脆弱双核阿米巴、结肠内阿米巴。

应用形态学AI识别的肠道寄生虫筛查检出率高，有效降低漏检风险。但单纯AI识别具有一定的假阳性，需经过人工复核。

To explore the application value of a fecal analysis instrument with morphological artificial intelligence (AI) recognition in the screening of intestinal parasites in feces.

A retrospective analysis was conducted using 9 845 fecal specimens collected from health examination participants in various towns of Tianmen City and received by the Laboratory Department of the First People′s Hospital of Tianmen from November 2023 to June 2025. Conventional manual microscopic examination and an AI based fecal analyzer were used to compare the detection rates of intestinal parasites. The consistency between the 2 methods was evaluated using the Kappa test. In addition, detection rates by sex, parasite species, and township distribution were analyzed.

The AI fecal analyzer identified a total of 294 positive samples for parasites, with a positive rate of 2.99% (294/9 845). After manual re-inspection, 80 cases were confirmed positive, with a positive rate of 0.81%. The false positive rate of AI identification was 2.19% (214/9 765). The positive rate of conventional manual microscopic examination was 0.56% (55/9 845), and the missed detection rate was 31.25% (25/80). There was a statistically significant difference in the detection rates between the two methods (χ²=21.330, P<0.01). There was no statistically significant difference in parasite detection rates between males and females for either the AI+ manual re-inspection method or the manual microscopic examination (χ²=0.109, 0.110, all P>0.05). The detection rates of parasites by simple AI identification and AI identification+ manual re-inspection reached a moderate consistency (Kappa=0.417, P<0.01), with a sensitivity of 98.75%, a specificity of 97.80%, a positive predictive value of 26.87%, and a negative predictive value of 99.99%. The detection rates of parasites by simple manual microscopic examination and AI identification+ manual re-inspection were highly consistent (Kappa=0.800, P<0.01), with a sensitivity of 67.50%, a specificity of 99.99%, a positive predictive value of 98.18%, and a negative predictive value of 99.73%. Both methods successfully detected the Clonorchis sinensis eggs, Giardia lamblia, pinworm eggs, and Blastocystis hominis, AI identification was also able to detect the Strongyloides stercoralis, Trichuris trichiura eggs, hookworm eggs, Entamoeba histolytica, and Entamoeba coli.

The morphology-based AI recognition system achieved a high detection rate for intestinal parasite screening and effectively reduced the risk of missed diagnoses. However, relying solely on AI recognition has a certain rate of false-positive results and therefore requiring manual verification.