Noninvasive Evaluation of the Pathologic Grade of Hepatocellular Carcinoma Using MCF-3DCNN: A Pilot Study

Abstract

Purpose: To evaluate the diagnostic performance of deep learning with a multichannel fusion three-dimensional convolutional neural network (MCF-3DCNN) in the differentiation of the pathologic grades of hepatocellular carcinoma (HCC) based on dynamic contrast-enhanced magnetic resonance images (DCE-MR images).

Methods and materials: Fifty-one histologically proven HCCs from 42 consecutive patients from January 2015 to September 2017 were included in this retrospective study. Pathologic examinations revealed nine well-differentiated (WD), 35 moderately differentiated (MD), and seven poorly differentiated (PD) HCCs. DCE-MR images with five phases were collected using a 3.0 Tesla MR scanner. The 4D-tensor representation was employed to organize the collected data in one temporal and three spatial dimensions by referring to the phases and 3D scanning slices of the DCE-MR images. A deep learning diagnosis model with MCF-3DCNN was proposed, and the structure of MCF-3DCNN was determined to approximate clinical diagnosis experience by taking into account the significance of the spatial and temporal information from DCE-MR images. Then, MCF-3DCNN was trained based on well-labeled samples of HCC lesions from real patient cases by experienced radiologists. The accuracy when differentiating the pathologic grades of HCC was calculated, and the performance of MCF-3DCNN in lesion diagnosis was assessed. Additionally, the areas under the receiver operating characteristic curves (AUC) for distinguishing WD, MD, and PD HCCs were calculated.

Results: MCF-3DCNN achieved an average accuracy of 0.7396±0.0104 with regard to totally differentiating the pathologic grade of HCC. MCF-3DCNN also achieved the highest diagnostic performance for discriminating WD HCCs from others, with an average AUC, accuracy, sensitivity, and specificity of 0.96, 91.00%, 96.88%, and 89.62%, respectively.

Conclusions: This study indicates that MCF-3DCNN can be a promising technology for evaluating the pathologic grade of HCC based on DCE-MR images.

Figure 1

Data representation with a 4th-order tensor.

Figure 2

The architecture of the MCF-3DCNN.

Figure 3

Axial MR images and pathologic image of a 70-year-old man with HCC. (a) A fat-suppressed T2-weighted fast spin-echo image shows an oval-shaped, slightly hyperintense neoplasm in the dorsal part of segment VI, with a maximum diameter of 2.6 cm. Axial precontrast (b), late artery phase (c), portal vein phase (d), equilibrium phase (e), and delay phase (f) T1-weighted 3D GRE images demonstrate a hypointense appearance of the lesion on precontrast T1-weighted images (b), high enhancement in the late arterial phase, (c) and washout in the portal vein phase (d) with an enhancing capsule that can clearly be observed in the equilibrium phase (e) and delay phase (f); all of these MRI features are consistent with typical HCC. The tumor was successfully surgically resected and was pathologically confirmed as a WD HCC.

Figure 4

Axial MR images and pathologic image of a 52-year-old man with HCC. (a) A fat-suppressed T2-weighted fast spin-echo image shows an oval-shaped, heterogeneous, slightly hyperintense neoplasm in segment VII with a maximum diameter of 4.6 cm. Axial precontrast (b), late artery phase (c), portal vein phase (d), equilibrium phase (e), and delay phase (f) T1-weighted 3D GRE images show a hyperintense appearance of the lesion on precontrast T1-weighted images (b), obvious enhancement in the late arterial phase (c), and washout in the portal vein phase (d); an enhancing capsule was detectable in the equilibrium phase (e) and was more obvious in the delay phase (f); all of these MRI features are consistent with typical HCC. The tumor was successfully surgically resected and was pathologically confirmed as an MD HCC with no vascular invasion.

Figure 5

Axial late arterial phase DCE-MR images of four pathologically confirmed PD HCCs. The imaging appearances in the arterial phase of different PD HCCs varied, presenting with homogeneous hyperenhancement (a), heterogeneous hyperenhancement with a small area of necrosis (b), heterogeneous hyperenhancement with a large area of necrosis (c), and hyperenhancement in some parts of tumor (d).

Figure 6

The average area under the ROC curve for 3DCNN for discriminating WD HCCs from the others was 0.96.

Figure 7

The average area under the ROC curve for 3DCNN for discriminating MD HCCs from the others was 0.71.

Figure 8

The average area under the ROC curve for 3DCNN for discriminating PD HCCs from the others was 0.64.