Sobek Joseph, Medina Inojosa Jose R, Medina Inojosa Betsy J, Rassoulinejad-Mousavi S M, Conte Gian Marco, Lopez-Jimenez Francisco, Erickson Bradley J
Department of Radiology, Mayo Clinic, Rochester, MN, USA.
Department of Cardiovascular Medicine, Mayo Clinic, Rochester, MN, USA.
J Imaging Inform Med. 2024 Dec;37(6):3208-3216. doi: 10.1007/s10278-024-01138-2. Epub 2024 Jun 6.
Artificial intelligence-enhanced identification of organs, lesions, and other structures in medical imaging is typically done using convolutional neural networks (CNNs) designed to make voxel-accurate segmentations of the region of interest. However, the labels required to train these CNNs are time-consuming to generate and require attention from subject matter experts to ensure quality. For tasks where voxel-level precision is not required, object detection models offer a viable alternative that can reduce annotation effort. Despite this potential application, there are few options for general-purpose object detection frameworks available for 3-D medical imaging. We report on MedYOLO, a 3-D object detection framework using the one-shot detection method of the YOLO family of models and designed for use with medical imaging. We tested this model on four different datasets: BRaTS, LIDC, an abdominal organ Computed tomography (CT) dataset, and an ECG-gated heart CT dataset. We found our models achieve high performance on a diverse range of structures even without hyperparameter tuning, reaching mean average precision (mAP) at intersection over union (IoU) 0.5 of 0.861 on BRaTS, 0.715 on the abdominal CT dataset, and 0.995 on the heart CT dataset. However, the models struggle with some structures, failing to converge on LIDC resulting in a mAP@0.5 of 0.0.
医学成像中人工智能增强的器官、病变和其他结构识别通常使用卷积神经网络（CNN）来完成，这些网络旨在对感兴趣区域进行体素精确分割。然而，训练这些CNN所需的标签生成耗时，并且需要主题专家的关注以确保质量。对于不需要体素级精度的任务，目标检测模型提供了一种可行的替代方案，可以减少标注工作量。尽管有这种潜在应用，但用于三维医学成像的通用目标检测框架选择很少。我们报告了MedYOLO，这是一种使用YOLO系列模型的一次性检测方法设计的三维目标检测框架，专为医学成像而设计。我们在四个不同的数据集上测试了这个模型：BRaTS、LIDC、一个腹部器官计算机断层扫描（CT）数据集和一个心电图门控心脏CT数据集。我们发现，即使不进行超参数调整，我们的模型在各种结构上也能实现高性能，在BRaTS上的交并比（IoU）为0.5时平均精度均值（mAP）达到0.861，在腹部CT数据集上为0.715，在心脏CT数据集上为0.995。然而，这些模型在一些结构上存在困难，在LIDC上无法收敛，导致mAP@0.5为0.0。
