经直肠超声引导下的系统活检是目前指南推荐的前列腺癌的确诊方法，但存在漏诊高级别前列腺癌和过度检出惰性前列腺癌的风险以及多针穿刺带来的高并发症风险问题。课题组前期研究证实基于超声图像分析的方法能够提取与组织微结构相关的特征参数，初步研究发现了前列腺癌灶与周围正常组织的超声图像特征参数存在显著差异。那么，能否通过影像组学的方法分析前列腺癌的特征参数从而实现癌灶可视化，指引靶向穿刺活检，目前尚未见报道。本研究拟通过经直肠前列腺三维超声容积图像与前列腺病理容积数据配准，在超声容积图像上对癌灶进行准确标识，采用深度卷积神经网络的方法构建基于三维超声容积图像的前列腺癌灶区域分割模型，实现癌灶在超声图像上可视化，达到经直肠超声引导下对前列腺癌靶向穿刺的目的，从而准确检出高级别前列腺癌，减少惰性前列腺癌的检出带来的过度治疗。研究对提高前列腺癌的治疗效果、减少系统穿刺活检针数以及相关的并发症均具有重要意义。

Transrectal ultrasound-guided systematic biopsy is currently recommended for the diagnosis of prostate cancer, but this approach is associated with significant morbidity and life-threatening sepsis, under-diagnosis of clinically significant prostate cancers and over-diagnosis of clinically insignificant cancers. Our previous study confirmed that analysis of ultrasound images could extract features of tissue microstructure. The preliminary study found that ultrasound features of prostate cancer were significantly different from those of the surrounding normal tissue. It has not been reported that whether this method, realizing the visualization of tumor location by radiomics, could guide target biopsy of prostate cancer. Our study intends to build a novel registration between transrectal ultrasound volume images of prostate and pathological volume data of prostate first, then mark the tumor location at transrectal ultrasound volume images, construct and validate a segmentation model of prostate cancer by deep convolutional neural network, finally achieve visualization of cancer location and realize target biopsy of prostate cancer by transrectal ultrasound. This method would be minimally invasive, cause fewer side effects, identify a high proportion of men who would benefit from treatment, and minimize the identification of men with clinically insignificant cancer in order to prevent overtreatment.