Registering histologic and MR images of prostate for image-based cancer detection

Abstract

Rationale and objectives: Needle biopsy is currently the only way to confirm prostate cancer. To increase prostate cancer diagnostic rate, needles are expected to be deployed at suspicious cancer locations. High-contrast magnetic resonance (MR) imaging provides a powerful tool for detecting suspicious cancerous tissues. To do this, MR appearances of cancerous tissue should be characterized and learned from a sufficient number of prostate MR images with known cancer information. However, ground-truth cancer information is only available in histologic images. Therefore it is necessary to warp ground-truth cancerous regions in histological images to MR images by a registration procedure. The objective of this article is to develop a registration technique for aligning histological and MR images of the same prostate.

Material and methods: Five pairs of histological and T2-weighted MR images of radical prostatectomy specimens are collected. For each pair, registration is guided by two sets of correspondences that can be reliably established on prostate boundaries and internal salient bloblike structures of histologic and MR images.

Results: Our developed registration method can accurately register histologic and MR images. It yields results comparable to manual registration, in terms of landmark distance and volume overlap. It also outperforms both affine registration and boundary-guided registration methods.

Conclusions: We have developed a novel method for deformable registration of histologic and MR images of the same prostate. Besides the collection of ground-truth cancer information in MR images, the method has other potential applications. An automatic, accurate registration of histologic and MR images actually builds a bridge between in vivo anatomical information and ex vivo pathologic information, which is valuable for various clinical studies.

Figure 1

Schematic description of our proposed computer-aided biopsy system. ➀ Generate optimal biopsy strategy based on patient-specific image information. ➁ Generate optimal biopsy strategy based on population-based statistical information. ➂ Integrate the two biopsy strategies and apply them to an individual patient.

Figure 2

An example of warping a ground-truth cancerous region from the histological image to the MR image of the same prostate. (a) Prostate histological image, where the dark pink region denotes ground-truth cancer labeled by a pathologist. (b) Prostate T2-weighted MR image. (c) Prostate T2-weighted MR image with manually warped cancer ground truth as indicated by a green region.

Figure 3

Geometric attributes of a boundary landmark. For a boundary landmark x_i, its geometric attributes are defined by the volumes of the tetrahedrons formed by vertices x_i and its neighbors nbr_l,0(x_i), nbr_l,m1(x_i), and nbr_l,m2(x_i). Here, m₁ = ⌊SI_l(x_i)/3⌋ and m₂ = ⌊SI_l(x_i)×2/3⌋, (⌊·⌋ defines the floor function) where SI_l(x_i) is the number of vertices contained by l-th neighborhood layer of x_i

Figure 4

Corresponding blob-like structures in prostate histological and MR images. (a) Prostate histological image. (b) Prostate MR images. Red arrows point to the corresponding blob-like structures commonly available in histological and MR images.

Figure 5

Schematic explanation of the scale-space analysis method. The local peak responses of normalized Laplacian describe important properties of blob-like structures and are used for selecting the candidates of internal landmarks.

Figure 6

Detection of internal landmarks. The internal landmarks are detected from prostate histological image (a) and MR image (b), respectively. The blue/red dots denote the centers of the detected blob-like structures and the sizes of the circles indicate the salient scales of the blob-like structures.

Figure 7

Correspondences between internal landmarks. The correspondences between internal landmarks in histological image (a) and MR images (b) are shown by color crosses. Crosses with the same color denote the corresponding internal landmarks.

Figure 8

Comparison of warping histological images to match with MR images by three different registration methods. Two red points and a red region in (a) denote the manually labeled landmarks and cancerous region in an MR image, respectively. For comparison, those red points and the boundary of cancerous region are repeatedly displayed in three warped histological images (b–d) by three registration methods, i.e., Methods 1, 2, and 3, respectively. The blue points in each of three warped histological images (b–d) are the warped landmarks manually labeled in original histological image, as correspondences to those red landmarks in MR image. The dark region in each warped histological image denotes the warped version of the manually labeled cancerous region in the histological image.