Deep‐learning based classification of a tumor marker for prognosis on Hodgkin's disease

Abstract

PurposeHodgkin's disease is a common malignant disorder in adolescent patients. Although most patients are cured, approximately 10%–15% of patients experience a relapse or have resistant disease. Furthermore, there are no definitive molecular predictors for early identification of patients at high risk of treatment failure to first line therapy. The aim of this study was to evaluate the deep learning‐based classifier model of medical image classification to predict clinical outcome that may help in appropriate therapeutic decisions.MethodsEighty‐three FFPE biopsy specimens from patients with Hodgkin's disease were stratified according to the patient's qPET scores, stained with picrosirius red dye and digitalized by whole slide image scanning. The resulting whole slide images were cut into tiles and annotated by two classes based on the collagen fibers' degree of coloring with picrosirius red. The neural network (YOLOv4) was then trained with the annotated data. Training was performed with 30 cases. Prognostic power of the weakly stained picrosirius red fibers was evaluated with 53 cases. The same neural network was trained with MMP9 stained tissue slides from the same cases and the quantification results were compared with the variant from the picrosirius red cases.ResultsThere was a weak monotonically increasing relationship by parametric ANOVA between the qPET groups and the percentages of weakly stained fibers (p = .0185). The qPET‐positive cases showed an average of 18% of weakly stained fibers, and the qPET‐negative cases 10%–14%. Detection performance showed an AUC of 0.79.ConclusionsPicrosirius red shows distinct associations as a prognostic metric candidate of disease progression in Hodgkin's disease cases using whole slide images but not sufficiently as a prognostic device.

Artificial intelligence in lymphoma histopathology: a systematic review (Preprint)

“…Several studies characterized the spatial location and relationships of typical cells and their surrounding cells, nuclei, and microenvironments of the regions of interest (17,28,32,38), demonstrating interpretability. One study presented graphical features associated with clinical prognostic information (33). A few studies employed traditional machine learning modeling, such as decision trees (21,25), which are inherently more understandable as they simulate human thought processes to make decisions.…”

“…Sharing code can help mitigate the impact of incomplete reporting and significantly improve reproducibility, but out of 31 papers, only five shared codes, some of which appeared to be incomplete or inaccessible. Better code repositories include detailed documentation to aid reproducibility, including information on environment setup, functionality overview, result production, and the code itself (10,12,15,21,33)： Some studies aim to provide interpretability for DL tools using current methods, including post hoc approaches or supervised ML models to interpret the outputs after DL models have made predictions (46,47). In the field of AI research for lymphoma, most studies currently provide personalized interpretability for analysis, including visual attention heatmaps, traditional ML showcasing the spatial location of feature areas, and relationships.…”