Automated Whole-Body Bone Lesion Detection for Multiple Myeloma on 68Ga-Pentixafor PET/CT Imaging Using Deep Learning Methods

Abstract

The identification of bone lesions is crucial in the diagnostic assessment of multiple myeloma (MM). ⁶⁸Ga-Pentixafor PET/CT can capture the abnormal molecular expression of CXCR-4 in addition to anatomical changes. However, whole-body detection of dozens of lesions on hybrid imaging is tedious and error prone. It is even more difficult to identify lesions with a large heterogeneity. This study employed deep learning methods to automatically combine characteristics of PET and CT for whole-body MM bone lesion detection in a 3D manner. Two convolutional neural networks (CNNs), V-Net and W-Net, were adopted to segment and detect the lesions. The feasibility of deep learning for lesion detection on ⁶⁸Ga-Pentixafor PET/CT was first verified on digital phantoms generated using realistic PET simulation methods. Then the proposed methods were evaluated on real ⁶⁸Ga-Pentixafor PET/CT scans of MM patients. The preliminary results showed that deep learning method can leverage multimodal information for spatial feature representation, and W-Net obtained the best result for segmentation and lesion detection. It also outperformed traditional machine learning methods such as random forest classifier (RF), k-Nearest Neighbors (k-NN), and support vector machine (SVM). The proof-of-concept study encourages further development of deep learning approach for MM lesion detection in population study.

Figure 1

Properties of MM lesions of an exemplary patient with ⁶⁸Ga-Pentixafor PET imaging: (a) maximum-intensity projection of ⁶⁸Ga-Pentixafor PET; (b) histogram distribution of maximum activity of the lesions; (c) histogram distribution of mean activity of the lesions; (d) histogram distribution of volumes of the lesions.

Figure 2

Overview of a simplified W-Net architecture.

Figure 3

Synthetic PET data generating from digital phantom and its corresponding CT scan.

Figure 4

Exemplary detection results of phantom study: (1) the original axial CT slices; (2) the corresponding PET slices; (3) MM lesion prediction using V-Net.

Figure 5

Exemplary detection results of V-Nets and W-Net: (1) the original axial CT slices; (2) the corresponding PET slices; (3) MM lesion prediction using CT alone in V-Net; (4) MM lesion prediction using PET alone in V-Net; (5) MM lesion prediction using PET/CT in V-Net; (6) MM lesion detection using W-Net.

Figure 6

Convergence curves of different network architectures, W-Net (a) or V-Net (b).