Machine learning-based pathomics signature of histology slides as a novel prognostic indicator in primary central nervous system lymphoma - Scorecard - MDSpire
Advertisement
Machine learning-based pathomics signature of histology slides as a novel prognostic indicator in primary central nervous system lymphoma
Clinical Scorecard: Prognostic Value of a Machine Learning-Derived Pathomics Signature from Histological Slides in Primary Central Nervous System Lymphoma
At a Glance
Category
Detail
Condition
Primary central nervous system lymphoma (PCNSL), a rare and aggressive extranodal non-Hodgkin lymphoma affecting CNS
Key Mechanisms
Machine learning-derived pathomics signature from H&E-stained histological slides using LASSO-Cox regression to predict prognosis
Target Population
Patients with histologically diagnosed CNS-DLBCL without systemic lymphoma
Care Setting
Specialized oncology and neuropathology centers with access to digital pathology and clinical follow-up
Key Highlights
PCNSL accounts for ~4% of CNS tumors and 4–6% of extranodal lymphomas, predominantly diffuse large B-cell lymphoma subtype
Existing prognostic models (IELSG and MSKCC) have limitations due to incomplete clinical data and inconsistent survival prediction
A novel automated pathomics score (Path-score) derived from digital histopathology images correlates with treatment response and improves prognostic accuracy
Guideline-Based Recommendations
Diagnosis
Histological diagnosis via H&E-stained slides remains gold standard for PCNSL
Exclude systemic lymphoma by CT or PET-CT and bone marrow aspiration
Collect comprehensive clinicopathological and follow-up data
Management
Initial treatment response assessment per International Primary CNS Lymphoma Collaborative Group criteria (CR, PR, SD, PD)
Use Path-score combined with clinical features in a nomogram to guide prognosis and potentially tailor therapy
Monitoring & Follow-up
Regular clinical follow-up to assess overall survival and treatment response
Monitor for relapse or progression given high rates of chemotherapy resistance and relapse
Risks
High risk of non-response (15–25%) and relapse (25–50%) after initial chemotherapy
Poor overall survival with 5-year OS rates between 22.3% and 35%
Patient & Prescribing Data
Patients with primary CNS-DLBCL without systemic involvement, confirmed by imaging and pathology
Path-score correlates with initial treatment response, distinguishing responders (CR/PR) from non-responders (SD/PD), aiding prognostication beyond traditional clinical scores
Clinical Best Practices
Incorporate digital pathology and automated feature extraction to complement traditional histopathological evaluation
Utilize machine learning models like LASSO-Cox regression to develop prognostic signatures from histological images
Combine novel pathomics scores with clinical parameters in nomograms for improved individualized outcome prediction
