1
|
Prezja F, Äyrämö S, Pölönen I, Ojala T, Lahtinen S, Ruusuvuori P, Kuopio T. Improved accuracy in colorectal cancer tissue decomposition through refinement of established deep learning solutions. Sci Rep 2023; 13:15879. [PMID: 37741820 PMCID: PMC10517936 DOI: 10.1038/s41598-023-42357-x] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/24/2023] [Accepted: 09/08/2023] [Indexed: 09/25/2023] Open
Abstract
Hematoxylin and eosin-stained biopsy slides are regularly available for colorectal cancer patients. These slides are often not used to define objective biomarkers for patient stratification and treatment selection. Standard biomarkers often pertain to costly and slow genetic tests. However, recent work has shown that relevant biomarkers can be extracted from these images using convolutional neural networks (CNNs). The CNN-based biomarkers predicted colorectal cancer patient outcomes comparably to gold standards. Extracting CNN-biomarkers is fast, automatic, and of minimal cost. CNN-based biomarkers rely on the ability of CNNs to recognize distinct tissue types from microscope whole slide images. The quality of these biomarkers (coined 'Deep Stroma') depends on the accuracy of CNNs in decomposing all relevant tissue classes. Improving tissue decomposition accuracy is essential for improving the prognostic potential of CNN-biomarkers. In this study, we implemented a novel training strategy to refine an established CNN model, which then surpassed all previous solutions . We obtained a 95.6% average accuracy in the external test set and 99.5% in the internal test set. Our approach reduced errors in biomarker-relevant classes, such as Lymphocytes, and was the first to include interpretability methods. These methods were used to better apprehend our model's limitations and capabilities.
Collapse
Affiliation(s)
- Fabi Prezja
- Faculty of Information Technology, University of Jyväskylä, Jyväskylä, 40014, Finland.
- Digital Health Intelligence Laboratory, University of Jyväskylä, Jyväskylä, 40014, Finland.
| | - Sami Äyrämö
- Faculty of Information Technology, University of Jyväskylä, Jyväskylä, 40014, Finland
- Digital Health Intelligence Laboratory, University of Jyväskylä, Jyväskylä, 40014, Finland
| | - Ilkka Pölönen
- Faculty of Information Technology, University of Jyväskylä, Jyväskylä, 40014, Finland
- Spectral Imaging Laboratory, University of Jyväskylä, Jyväskylä, 40014, Finland
| | - Timo Ojala
- Faculty of Information Technology, University of Jyväskylä, Jyväskylä, 40014, Finland
- Digital Health Intelligence Laboratory, University of Jyväskylä, Jyväskylä, 40014, Finland
| | - Suvi Lahtinen
- Faculty of Information Technology, University of Jyväskylä, Jyväskylä, 40014, Finland
- Department of Biological and Environmental Science, Faculty of Mathematics and Science, University of Jyväskylä, Jyväskylä, 40014, Finland
| | - Pekka Ruusuvuori
- Institute of Biomedicine, Cancer Research Unit, University of Turku, Turku, 20014, Finland
- FICAN West Cancer Centre, Turku University Hospital, Turku, 20521, Finland
| | - Teijo Kuopio
- Department of Education and Research, Hospital Nova of Central Finland, Jyväskylä, 40620, Finland
- Department of Biological and Environmental Science, University of Jyväskylä, Jyväskylä, 40014, Finland
- Department of Pathology, Hospital Nova of Central Finland, Jyväskylä, 40620, Finland
| |
Collapse
|
2
|
Khan U, Koivukoski S, Valkonen M, Latonen L, Ruusuvuori P. The effect of neural network architecture on virtual H&E staining: Systematic assessment of histological feasibility. PATTERNS (NEW YORK, N.Y.) 2023; 4:100725. [PMID: 37223268 PMCID: PMC10201298 DOI: 10.1016/j.patter.2023.100725] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 10/27/2022] [Revised: 12/23/2022] [Accepted: 03/08/2023] [Indexed: 05/25/2023]
Abstract
Conventional histopathology has relied on chemical staining for over a century. The staining process makes tissue sections visible to the human eye through a tedious and labor-intensive procedure that alters the tissue irreversibly, preventing repeated use of the sample. Deep learning-based virtual staining can potentially alleviate these shortcomings. Here, we used standard brightfield microscopy on unstained tissue sections and studied the impact of increased network capacity on the resulting virtually stained H&E images. Using the generative adversarial neural network model pix2pix as a baseline, we observed that replacing simple convolutions with dense convolution units increased the structural similarity score, peak signal-to-noise ratio, and nuclei reproduction accuracy. We also demonstrated highly accurate reproduction of histology, especially with increased network capacity, and demonstrated applicability to several tissues. We show that network architecture optimization can improve the image translation accuracy of virtual H&E staining, highlighting the potential of virtual staining in streamlining histopathological analysis.
Collapse
Affiliation(s)
- Umair Khan
- University of Turku, Institute of Biomedicine, Turku 20014, Finland
| | - Sonja Koivukoski
- University of Eastern Finland, Institute of Biomedicine, Kuopio 70211, Finland
| | - Mira Valkonen
- Tampere University, Faculty of Medicine and Health Technology, Tampere 33100, Finland
| | - Leena Latonen
- University of Eastern Finland, Institute of Biomedicine, Kuopio 70211, Finland
- Foundation for the Finnish Cancer Institute, Helsinki 00290, Finland
| | - Pekka Ruusuvuori
- University of Turku, Institute of Biomedicine, Turku 20014, Finland
- Tampere University, Faculty of Medicine and Health Technology, Tampere 33100, Finland
- FICAN West Cancer Centre, Cancer Research Unit, Turku University Hospital, Turku 20500, Finland
| |
Collapse
|