Machine learning-based pathomics signature of histology slides as a novel prognostic indicator in primary central nervous system lymphoma

Ling Duan ¹, Yongqi He ¹, Wenhui Guo ¹

⁰Department of Neuro-Oncology, Cancer Center, Beijing Tiantan Hospital, Capital Medical University, No.119 West Nansihuan Road, Beijing, 100070, China.

Journal of Neuro-Oncology +

|

April 1, 2024

View abstract on PubMed

Summary

A new pathomics signature accurately predicts outcomes for Primary Central Nervous System Lymphoma (PCNSL). This non-invasive approach offers a powerful tool for assessing overall survival and progression-free survival in PCNSL patients.

Area Of Science

Digital pathology
Computational oncology
Biomarker discovery

Background

Primary Central Nervous System Lymphoma (PCNSL) is a rare and aggressive brain tumor.
Accurate prediction of patient outcomes and therapeutic response is crucial for effective management.
Existing prognostic models for PCNSL have limitations.

Purpose Of The Study

To develop and validate a novel pathomics signature for predicting outcomes in PCNSL.
To assess the signature's ability to predict overall survival (OS) and progression-free survival (PFS).
To evaluate the signature's potential as a predictive indicator for treatment response.

Main Methods

Quantitative features were extracted from hematoxylin and eosin (H&E) stained whole-slide images (WSIs) of 114 PCNSL patients using CellProfiler.
A pathomics signature was developed and validated using Cox regression, ROC curves, Calibration, DCA, and NRI.
Machine learning classifiers were employed to analyze the extracted features.

Main Results

A total of 802 quantitative features were extracted via an automated pipeline.
The pathomics signature significantly predicted OS and PFS in both training and validation cohorts.
Patients with a high Path-score exhibited a significantly lower response rate to initial treatment compared to those with a low Path-score.
The developed nomogram demonstrated incremental performance over existing models with high AUC values.

Conclusions

The developed pathomics signature is a powerful, non-invasive predictor of OS and PFS in PCNSL.
This novel approach may serve as a valuable predictive indicator for therapeutic response in PCNSL.
Pathomics offers a convenient and effective tool for improving PCNSL patient management.

Keywords:

Histopathological images Pathomics Primary central nervous system lymphoma Prognosis