Prediction of Clinical Outcomes with Explainable Artificial Intelligence in Patients with Chronic Lymphocytic Leukemia

Abstract

Background: The International Prognostic Index (IPI) is applied to predict the outcome of chronic lymphocytic leukemia (CLL) with five prognostic factors, including genetic analysis. We investigated whether multiparameter flow cytometry (MPFC) data of CLL samples could predict the outcome by methods of explainable artificial intelligence (XAI). Further, XAI should explain the results based on distinctive cell populations in MPFC dot plots.

Methods: We analyzed MPFC data from the peripheral blood of 157 patients with CLL. The ALPODS XAI algorithm was used to identify cell populations that were predictive of inferior outcomes (death, failure of first-line treatment). The diagnostic ability of each XAI population was evaluated with receiver operating characteristic (ROC) curves.

Results: ALPODS defined 17 populations with higher ability than the CLL-IPI to classify clinical outcomes (ROC: area under curve (AUC) 0.95 vs. 0.78). The best single classifier was an XAI population consisting of CD4+ T cells (AUC 0.78; 95% CI 0.70-0.86; p < 0.0001). Patients with low CD4+ T cells had an inferior outcome. The addition of the CD4+ T-cell population enhanced the predictive ability of the CLL-IPI (AUC 0.83; 95% CI 0.77-0.90; p < 0.0001).

Conclusions: The ALPODS XAI algorithm detected highly predictive cell populations in CLL that may be able to refine conventional prognostic scores such as IPI.

Keywords: ALPODS; artificial intelligence; chronic lymphocytic leukemia; flow cytometry.

Figure 1

Workflow and data processing. Flow cytometry raw data was standardized and assigned to the outcome group (TTF 1 inferior, TTF 0 superior). The ALPODS algorithm was used to identify distinctive cell populations with different frequencies in TTF 1 versus TTF 0 The most important populations for determination were visualized in flow cytometry bivariate dot plots and assigned to their biological counterparts.

Figure 2

Multiple logistic regression with ROC curve analysis for (A) all XAIpopulations revealed a predictive ability of AUC 0.95 (95% CI 0.91–0.98; p < 0.0001). (B): The restriction of the four most predictive XAI populations for the outcome (TTF) and prognosis (IPI) resulted in lower predictive ability (AUC 0.87; 95% CI 0.80–0.93; p < 0.0001). Abbreviations: ROC = receiver operation characteristics; XAI = explainable artificial intelligence; AUC = area under curve; CI = confidence interval.

Figure 3

Identification and description of the XAI population through flow cytometry gating in a representative sample. (A) The populations T1C0011 (green) and T1C0016 (red) were the most relevant populations for the outcome and prognosis in Tube 1 (T1) of the analyzed B-cell panel. T1C0011 (green) could be located within the CLL cells (blue). The population of T1C0016 (red) corresponded to CD4+ T cells (i.e., T helper cells). (B) T2C0004 (green) and T2C0018 (red) were the most relevant populations for the outcome and prognosis in Tube 2 (T2). Both were located mainly within the CLL cells (blue), but T2C0018 was a mixture of a biologically different population (CLL cells and T/NK cells).

Figure 4

Median antigen expression and scatter properties of CLL subsets with predictive ability for outcome (TTF) were compared with the CLL cells of the average patient. The population T1C0011 (A), which was identified in tube 1 (T1), showed lower forward scatter (FS) and diminished antigen expression. T2C0002 (B) in tube 2 (T2) shared the lower FS and antigen expression with T1C0011. In flow cytometry dot plots (C), parts of T1C0011 and T2C0002 were located in the very low FS region of apoptotic/dead CLL cells. This may indicate a predictive value of apoptotic/dead CLL cells for outcome.