Computational approach for counting of SISH amplification signals for HER2 status assessment

View article
PeerJ Computer Science

Main article text

 

Introduction

  1. Introducing a robust methodology for automated detection and scoring of HER2/CEP17 biomarkers from identified nuclei in SISH stain images, addressing the practical need for accurate and reproducible HER2 status assessment.

  2. Evaluating the concordance between computer-aided HER2 status determination and expert pathologist assessments using SISH, thus demonstrating the improvement of the proposed method over current practices.

  3. Investigating factors influencing inter-observer interpretative reproducibility among pathologists, providing insights into the practical limitations and potential of the technique.

Background Study

Material and Methodology

Material

System overview

Preprocessing

  • xi is the normalized value of xi,

  • min(x) is the minimum value in array x,

  • max(x) is the maximum value in array x,

  • lower_bound is the lower bound of the desired range,

  • upper_bound is the upper bound of the desired range.

HER2 and CEP17 signal detection

Nuclei detection

  • Identification of nuclei with at least two CEP17 signals.

  • Exclusion of nuclei with missing or excessively overlapped information.

  • Polygon representation: Each nucleus is represented as a polygon, making it easier to handle overlaps and partial detections.

  • Robust training: The model was fine-tuned using a dataset annotated by experts, ensuring high accuracy in nuclei detection.

  • Efficient processing: By decomposing large image patches into smaller segments (256 × 256), the model maintains high performance while reducing computational load.

  1. Correlation coefficient: Measures the correlation between the automated nuclei count and manual enumeration by pathologists. A high correlation coefficient (0.91 in our case) indicates that the model’s predictions closely match expert assessments.

  2. Statistical significance (p-value): a one-sided paired t-test was conducted to compare the differences between the outcomes of the proposed method and the reference standard. p-values exceeding 0.05 indicated no significant statistical difference, validating the reliability of our approach.

Signal quantification

Experimental Setup

Parameter setting

Hardware specifications

  • Operating System: Windows 11 (64 bit)

  • CPU: 13th Gen Intel(R) Core(TM) i7-13620H @ 2.40 GHz

  • RAM: 32 GB

  • Graphics Card RAM Size: 8 GB

Evaluation criteria

  • The overall agreement between the manual and automated methods.

  • Any systematic bias in the automated method.

  • The consistency and reliability of the automated method across different samples.

Visual results

Quantification results

  • Positive: HER2/CEP17 ration <2.0 or HER2 copy number ≥6.0 signals per cell.

  • Negative: HER2/CEP17 ratio <2.0 and HER2 copy number <4.0 signals per cell.

  • Equivocal: HER2/CEP17 ratio <2.0, but HER2 copy number is 4.0–6.0 signals per cell.

Discussion

Advantages and limitations

Conclusion

Supplemental Information

Code for the whole experiment conducted for analysis purpose

DOI: 10.7717/peerj-cs.2373/supp-1

Deep learning model that use for Nuclei segmentation

DOI: 10.7717/peerj-cs.2373/supp-2

Additional Information and Declarations

Competing Interests

The authors declare there are no competing interests.

Author Contributions

Zaka Ur Rehman conceived and designed the experiments, performed the experiments, analyzed the data, performed the computation work, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Mohammad Faizal Ahmad Fauzi conceived and designed the experiments, authored or reviewed drafts of the article, and approved the final draft.

Wan Siti Halimatul Munirah Wan Ahmad performed the experiments, analyzed the data, authored or reviewed drafts of the article, and approved the final draft.

Fazly Salleh Abas conceived and designed the experiments, authored or reviewed drafts of the article, and approved the final draft.

Phaik Leng Cheah conceived and designed the experiments, authored or reviewed drafts of the article, data annotation, and approved the final draft.

Seow Fan Chiew conceived and designed the experiments, authored or reviewed drafts of the article, data curation, and approved the final draft.

Lai-Meng Looi performed the experiments, authored or reviewed drafts of the article, data Validation, and approved the final draft.

Ethics

The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers):

Department of Pathology, University of Malaya Medical Centre’s Medical Research Ethics Committee (UMMCMREC) (Ref MREC-ID 202195-10555).

Data Availability

The following information was supplied regarding data availability:

The data is available at figshare: Rehman, Zaka Ur (2024). Expert Annotated SISH Images Patches for Nuclei Segmentation. figshare. Figure. https://doi.org/10.6084/m9.figshare.26181908.v1.

This database is from the Hospital and is not public to protect patient confidentiality. Access to the data must be requested from the Chair of the University of Malaya Medical Centre Medical Research Ethics Committee (MREC): ummc-mrec@ummc.edu.my.

Funding

This work was supported by the Fundamental Research Grant Scheme (FRGS), Malaysia (FRGS/1/2020/ICT02/MMU/02/10). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

1 Citation 443 Views 9 Downloads