The Difference Between Radiomic Print of Hodgkin and Non- Hodgkin Lymphoma | ||||
| SECI Oncology Journal | ||||
| Volume 13, Issue 1, January 2025, Page 78-83 PDF (395.25 K) | ||||
 View on SCiNiTO | ||||
| Abstract | ||||
| Background: Lymphoma is one of the leading causes of death in adults and children. Diagnosing a specific subclass of lymphoma requires comprehensive histological evaluation and immune histochemistry analysis. In this study, we examine the ability of an artificial intelligence-based classifier to differentiate between Hodgkin disease (HD) and non-Hodgkin lymphoma (NHL) based on their radiomic print. Methods: A retrospective cohort was conducted for the patients diagnosed with lymphoma between 2019 and 2023. The initial baseline PETCT scans of these patients were retrieved. The active lesions were segmented and the radiomic features were extracted. The collected features were split into a training set (80%) and a validation set (20%). The primary endpoint of this study was used to build a classifier that could predict the type of lymphoma (Hodgkin or Non- Hodgkin). The training set was used to develop the model and the validation set was used to validate the results. Results: The study included 78 patients. Hodgkin disease was seen in 51 patients. The total number of identified and segmented lesions was 222, and 111 of them were retrieved from HD scans. Radiomic features were extracted from the PETCT. Several modelling approaches were examined. The highest accuracy was seen with the TabPFN classifier with a validation set accuracy of 73.3%. The model achieved an F1-score of 0.76 and 0.70 for HD and NHL, respectively. Conclusion: The TabPFN-based classifier achieved an accuracy of 73.3% on the validation sets. Further research on large sets is necessary. | ||||
| Keywords | ||||
| Radiomics; radiomics in lymphoma; difference between Hodgkin and non-Hodgkin lymphoma; AI in lymphoma | ||||
|
Statistics Article View: 68 PDF Download: 63 |
||||
