Villamizar-Monsalve María Alejandra, Sánchez-Montejo Javier, López-Abán Julio, Vicente Belén, Marín Miguel, Fernández-Ceballos Noelia, Peláez Rafael, Muro Antonio
Infectious and Tropical Diseases Research Group (e-INTRO), Biomedical Research Institute of Salamanca, Research Centre for Tropical Diseases at the University of Salamanca (IBSAL-CIETUS), Faculty of Pharmacy, University of Salamanca, 37007 Salamanca, Spain.
Organic and Pharmaceutical Chemistry Department, Biomedical Research Institute of Salamanca Research (IBSAL), Faculty of Pharmacy, University of Salamanca, 37007 Salamanca, Spain.
Biomedicines. 2024 Dec 19;12(12):2894. doi: 10.3390/biomedicines12122894.
Schistosomiasis impacts over 230 million people globally, with 251.4 million needing treatment. The disease causes intestinal and urinary symptoms, such as hepatic fibrosis, hepatomegaly, splenomegaly, and bladder calcifications. While praziquantel (PZQ) is the primary treatment, its effectiveness against juvenile stages (schistosomula) is limited, highlighting the need for new therapeutic agents, repurposed drugs, or reformulated compounds. Existing microscopy methods for assessing schistosomula viability are labor-intensive, subjective, and time-consuming.
An artificial intelligence (AI)-assisted culture system using YOLOv5 was developed to evaluate compounds against schistosomula. The AI model, based on object detection, was trained on 4390 images distinguishing between healthy and damaged schistosomula. The system was externally validated against human counters, and a small-scale assay was performed to demonstrate its potential for larger-scale assays in the future.
The AI model exhibited high accuracy, achieving a mean average precision (mAP) of 0.966 (96.6%) and effectively differentiating between healthy and damaged schistosomula. External validation demonstrated significantly improved accuracy and counting time compared to human counters. A small-scale assay was conducted to validate the system, identifying 28 potential compounds with schistosomicidal activity against schistosomula in vitro and providing their preliminary LC values.
This AI-powered method significantly improves accuracy and time efficiency compared to traditional microscopy. It enables the evaluation of compounds for potential schistosomiasis drugs without the need for dyes or specialized equipment, facilitating more efficient drug assessment.
血吸虫病在全球影响超过2.3亿人，有2.514亿人需要治疗。该疾病会引发肠道和泌尿系统症状，如肝纤维化、肝肿大、脾肿大和膀胱钙化。虽然吡喹酮（PZQ）是主要治疗药物，但其对幼虫阶段（童虫）的有效性有限，这凸显了对新型治疗药物、老药新用或重新配方化合物的需求。现有的评估童虫活力的显微镜检查方法劳动强度大、主观且耗时。
开发了一种使用YOLOv5的人工智能（AI）辅助培养系统，以评估针对童虫的化合物。该基于目标检测的AI模型在4390张区分健康和受损童虫的图像上进行训练。该系统与人工计数进行了外部验证，并进行了小规模试验以证明其未来用于大规模试验的潜力。
AI模型表现出高准确性，平均精度均值（mAP）达到0.966（96.6%），并能有效区分健康和受损童虫。外部验证表明，与人工计数相比，准确性显著提高，计数时间缩短。进行了小规模试验以验证该系统，确定了28种在体外对童虫具有杀血吸虫活性的潜在化合物，并提供了它们的初步半数致死浓度（LC）值。
与传统显微镜检查相比，这种基于AI的方法显著提高了准确性和时间效率。它能够在无需染料或专门设备的情况下评估化合物作为潜在血吸虫病药物的效果，有助于更高效地进行药物评估。
