Artificial Intelligence-Based Flow Cytometry Accurately Classifies, Prognosticates MDS

Researchers have developed an artificial-intelligence (AI)-based multiparametric flow cytometry (MFC) score that can accurately classify and risk stratify myelodysplastic syndromes (MDS) compared to the 2022 Molecular International Prognostic Scoring System (IPSS-M). This MFC score has allowed MDS to be diagnosed with a sensitivity of 91.8% and a specificity of 92.5%.

The 2022 IPSS-M classifies patients into six risk categories by looking at hematological parameters, cytogenetic abnormalities, and somatic mutations in 31 genes. However, the score does not include flow cytometry data, the researchers noted.

Valentin Clichet, of the Hôpital Saint-Louis APHP; and Thomas Boyer of Centre Hospitalier Universitaire d’Amiens, outlined their comparative analysis in a letter to the editor published in Haematologica.

IPSS-M

Using data from 119 patients with complete molecular and MFC characteristics, the researchers compared their MFC score with the IPSS-M score and observed a P value of <.001. Based on the IPSS-M, 44 (37%) patients had low-risk score, 17 (14%) had moderate high, 16 (13.5%) had moderate low, 16 (13.5%) had very low, 14 (12%) had high, and 12 (10%) had very high.

The researchers observed a mean of 2.02 for the MDS score and a mean of -0.30 for the IPSS-M score. For the AI-based MDS score, based on the six IPSS-M categories, they observed a mean of 0.8 for the very low group, a mean of 0.82 for the low group, a mean of 1.69 for the moderate low group, a mean of 2.41 for the moderate high group, a mean of 4.6 for the high group, and a mean of 4.9 for the very high group.

“The results…clearly show a significant positive association between the AI-based MDS score and the IPSS-M score,” the researchers wrote.

World Health Organization (WHO)

Another comparison between the WHO 2022 groups and the AI-based MDS score indicated a significant difference (P<.001). Based on WHO classifications, 55 (46%) patients had MDS with low blast, 21 (18%) had increased blasts 1, 18 (15%) had SF3B1 mutation, 12 (20%) had increased blasts 2, eight (7%) had biallelic TP53 inactivation, and five (4%) had 5q deletion.

According to post-HOC tests, there were significant differences between classifications of the MDS increased blasts 1 and low blast groups (3.39 vs 0.82, P=.001), the increased blasts 2 and low blast (4.1 vs 0.82, P <.001), the biallelic TP53 inactivation and low blast (6.18 vs 0.82, P<.001), the increased blasts 2 and the SF3B1 mutated (4.1 vs 1.25, P=.03), the biallelic TP53 inactivation and SF3B1 mutated (6.18 vs 1.25, P<.001), and the biallelic TP53 inactivation and the 5q deletion groups (6.18 vs 0.69, P=.002).

International Consensus Classification (ICC)

A comparison between the ICC 2022 and the MFC score showed another significant difference (P<.001). Based on ICC classifications, 29 (24%) patients had MDS with multilineage dysplasia, 21 (18%) had excess blasts, 18 (15%) had mutated SF3B1, 15 (13%) were not otherwise specified, 11 (9%) had single lineage dysplasia, 11 (9%) had an excess of blasts >10% in bone marrow representing the MDS/acute myeloid leukemia (AML) category, nine (8%) had multi-hit TP53, and five (4%) had 5q deletion.

The most significant differences were between MDS single lineage dysplasia and MDS/AML TP53 (1.7 vs 9.5, P<.001), the 5q deletion and MDS/AML TP53 (0.7 vs 9.5, P<.001), MDS not otherwise specified and MDS/AML (0.1 vs 4.3, P=.001), MDS excess blasts and MDS/AML TP53 (3.4 vs 9.5, P=.003), and between MDS multilineage dysplasia and MDS/AML TP53 groups (0.84 vs 9.5, P=.003).

In conclusion, “we compared the AI-based MDS score’s performance with the latest classifications of hemopathies and the IPSS-M prognostic score and found a perfect correlation between the score and these different entities,” the researchers wrote. Study limitations included the cohort size, as patient data was taken from three hospitals.

Reference

Clichet V, Boyer T. Artificial intelligence-based Myelodysplastic Syndromes Score, 2022 classifications, and the Molecular International Prognostic Scoring System : a perfect match. Haematologica. 2024. doi:10.3324/haematol.2024.286340