Evaluating ADC-only deep learning pipelines for breast cancer detection and segmentation using standalone diffusion-weighted MRI
Authors: Pablo García Marcos, Paula Puerta Gonzĺez, Guillermo Lorenzo, Héctor Gómez, Covadonga del Camino, Angel Rio-Alvarez, Víctor M. González
Organizations: Computer Sciences Department, University of Oviedo, Asturias, Spain · Biomedical Engineering Center (BME), University of Oviedo, Asturias, Spain · Electrical Engineering Department, University of Oviedo, Asturias, Spain · Group of Numerical Methods in Engineering, Department of Mathematics, University of A Coruña, A Coruña, Spain · Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin, Austin, Texas, USA · School of Mechanical Engineering, Purdue University, West Lafayette, Indiana, USA · Weldon School of Biomedical Engineering, Purdue University, West Lafayette, Indiana, USA · Purdue Center for Cancer Research, Purdue University, West Lafayette, Indiana, USA · Radiodiagnostic Service. Asturias Central University Hospital (HUCA). Oviedo, Spain
Dynamic contrast-enhanced (DCE) imaging is the gold standard technique for the detection and characterization of breast cancer using magnetic resonance imaging (MRI). However, DCE-MRI requires long acquisition times and the administration of contrast into the bloodstream, which can cause allergic reactions. Alternatively, diffusion-weighted MRI (DW-MRI) is a standard complementary technique for breast MRI that does not require contrast, has shorter acquisition times, and enables calculation of apparent diffusion coefficient (ADC) maps that correlate with tumor cellularity. Yet, despite these technical advantages, deep learning research has focused on DCE-based models and has barely explored the tumor detection performance of DW-MRI and ADC maps either in combination with DCE-MRI or as standalone alternatives. Here, we evaluate the application of different state-of-the-art deep learning techniques for detection and segmentation of breast cancer using ADC-only images. This is, to our knowledge, the first comprehensive evaluation of ADC-only breast cancer pipelines for classification, detection, and segmentation tasks.
Figures & tables
Figure 1: Slice of a study of the dataset in DW (left), ADC (center) and mask (right) formats.
Fold
Precision
Recall
Accuracy
Specificity
1
0.945
0.896
0.933
0.960
2
0.958
0.865
0.921
0.969
3
0.954
0.823
0.910
0.971
4
0.916
0.813
0.881
0.938
5
0.939
0.826
0.905
0.962
Mean
0.942
0.844
0.910
0.960
Table 1: Performance of the model in slice classification.
Figure 2: Classification results: ADC slices correctly classified by the model.
Figure 3: Classification results: ADC slices incorrectly classified by the model.
Fold
Precision
Recall
AP
1
0.917
0.767
0.840
2
0.894
0.671
0.775
3
0.864
0.634
0.722
4
0.842
0.649
0.730
5
0.878
0.681
0.724
Mean
0.878
0.680
0.758
Table 2: Performance of the model in tumor detection and localization.
Model
DICE
2D
0.699
3D
0.696
Table 3: Performance in breast cancer tumor semantic segmentation.
Figure 4: Localization results: (a) The ground truth shows the tumor enclosed by a blue box and assigned the label 0. (b) The tumor is enclosed in a blue box labeled as class 0, with the confidence score (0.9) shown next to the label.
Figure 5: Localization results: (a) The ground truth shows the tumor enclosed by a blue box and assigned the label 0. (b) The model prediction shows two boxes with label 0, one with confidence score 0.9 (correct) and one with confidence score 0.6 (incorrect).
German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Germany · Medical Faculty, Heidelberg University, Germany · Faculty of Mathematics and Computer Science, Heidelberg University, Germany +3
Institute of Radiology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Maximiliansplatz 3, 91054 Erlangen, Germany · Institute of Computer Science, Polish Academy of Science, Warsaw, Poland · Department of Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany +2
German Cancer Research Center (DKFZ) Heidelberg, Division of Medical Image Computing, Germany · Medical Faculty, Heidelberg University, Germany · Faculty of Mathematics and Computer Science, Heidelberg University, Germany +3