SBIR Phase I: AI-based system for analyzing multiparametric MRI scans for prostate lesion detection

小企业创新研究(SBIR)第一阶段项目的更广泛影响/商业潜力将是改善前列腺癌诊断。该项目提出了一个基于人工智能(AI)的平台，克服了目前人工磁共振成像(MRI)评估的缺点，目前在高质量的机器上进行，结果由全国几个地点的专家放射科医生读取。人工智能使MRI分析实现了自动化、标准化和更准确。这将减少不必要的侵入性活检和/或低风险前列腺癌的治疗，以及医生运行和读取核磁共振成像的时间和精力。此外，它将使非专家中心能够准确地诊断患者，而不需要广泛的测试、顶级核磁共振设备或侵入性手术。这个小型企业创新研究(SBIR)第一阶段项目开发了一个新的前列腺癌诊断平台，该平台利用基于人工智能的图像分析的强大功能来实现高灵敏度和特异度。虽然前列腺癌的诊断在很大程度上依赖于磁共振成像，但这些评估的准确性取决于专家经验和磁共振成像质量。机器学习解决方案能够对检测到的具有临床意义的病变的可能性进行全球评估(即Gleason 3+4/2级或更高级别)，更好地向临床医生提供适当的治疗信息。该项目提出了以下技术开发目标：1)开发一种测量MRI质量的方法，2)创建一个生成性对抗网络(GAN)框架，以修复从较差到较高质量的图像，以及3)前瞻性地评估新框架，以衡量AI算法的整体性能。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of the Small Business Innovation Research (SBIR) Phase I project will be to improve prostate cancer diagnostics. This project proposes an artificial intelligence (AI)-based platform that overcomes current shortcomings with manual magnetic resonance imaging (MRI) assessment, currently conducted on high-quality machines with results read by expert radiologists at a few locations around the country. AI enables MRI analysis that is automated, standardized, and more accurate. This will reduce unnecessary and invasive biopsy and/or treatment for low-risk prostate cancers, as well as physician time and effort to run and read MRIs. Furthermore, it will enable non-expert centers to accurately diagnose patients without requiring extensive testing, top-tier MRI equipment, or invasive surgery. This Small Business Innovation Research (SBIR) Phase I project develops a novel prostate cancer diagnostic platform that leverages the power of AI-based image analysis for high sensitivity and specificity. While prostate cancer diagnostics rely heavily on MRIs, the accuracy of these assessments depends on both expert experience and MRI quality. The machine learning solution is able to make global assessments on the likelihood of detected lesions being clinically significant (i.e., Gleason 3+4/grade group 2 or higher), better informing clinicians on appropriate treatment. This project proposes the following technology development objectives: 1) Develop a method to measure MRI quality, 2) Create a Generative Adversarial Network (GAN) Framework to repair images from poor to higher quality, and 3) Prospectively evaluate new framework to measure the overall performance of the AI algorithm.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.