School of Engineering and Design, Brunel University West London, Uxbridge, Middlesex, UB8 3PH, UK.
Int J Comput Assist Radiol Surg. 2013 Jul;8(4):547-60. doi: 10.1007/s11548-013-0813-y. Epub 2013 Jan 25.
To improve the computer-aided diagnosis of breast lesions, by designing a pattern recognition system (PR-system) on commercial graphics processing unit (GPU) cards using parallel programming and textural information from multimodality imaging.
Patients with histologically verified breast lesions underwent both ultrasound (US) and digital mammography (DM), lesions were outlined on the images by an experienced radiologist, and textural features were calculated. The PR-system was designed to provide highest possible precision by programming in parallel the multiprocessors of the NVIDIA's GPU cards, GeForce 8800GT or 580GTX, and using the CUDA programming framework and C++. The PR-system was built around the probabilistic neural network classifier, and its performance was evaluated by a re-substitution method, for estimating the system's highest accuracy, and by the external cross-validation method, for assessing the PR-system's unbiased accuracy to new, "unseen" by the system, data.
Classification accuracies for discriminating malignant from benign lesions were as follows: 85.5 % using US-features alone, 82.3 % employing DM features alone, and 93.5 % combining US and DM features. Mean accuracy to new "unseen" data for the combined US and DM features was 81 %. Those classification accuracies were about 10 % higher than accuracies achieved on a single CPU, using sequential programming methods, and 150-fold faster.
The proposed PR-system improves breast-lesion discrimination accuracy, it may be redesigned on site when new verified data are incorporated in its depository, and it may serve as a second opinion tool in a clinical environment.
通过使用并行编程和多模态成像的纹理信息，在商业图形处理单元 (GPU) 卡上设计模式识别系统 (PR 系统)，提高对乳腺病变的计算机辅助诊断。
经组织学证实患有乳腺病变的患者接受超声 (US) 和数字乳腺摄影 (DM) 检查，由有经验的放射科医生在图像上勾画病变，并计算纹理特征。该 PR 系统旨在通过并行编程 NVIDIA 的 GPU 卡，GeForce 8800GT 或 580GTX 的多核处理器，以及使用 CUDA 编程框架和 C++，实现最高可能的精度。PR 系统围绕概率神经网络分类器构建，通过重新替换方法评估系统的最高精度，通过外部交叉验证方法评估 PR 系统对新的、系统未见过的数据的无偏精度。
单独使用 US 特征区分良恶性病变的分类准确率为 85.5%，单独使用 DM 特征为 82.3%，联合使用 US 和 DM 特征为 93.5%。联合使用 US 和 DM 特征对新的“未见过”数据的平均准确率为 81%。与使用顺序编程方法在单个 CPU 上实现的准确率相比，这些分类准确率高出约 10%，速度提高了 150 倍。
所提出的 PR 系统提高了乳腺病变的鉴别准确率，当新的验证数据纳入其存储库时，它可以重新设计，并且可以作为临床环境中的辅助诊断工具。
