Semiautomatic assessment of immunofluorescence microscopy on blood smears in inherited platelet disorders using artificial intelligence: a proof of concept.

Introduction Inherited platelet disorders (IPD) are a group of rare diseases caused by mutations in more than 70 genes. Despite advances in molecular diagnosis, platelet phenotyping still represents a cornerstone in the diagnostic workup of IPD. Immunofluorescence microscopy on the blood smear, combined with light microscopy, has been established as a screening tool for IPD and validated as sensitive and specific for 9 of the most frequent IPD [1]. A major drawback of the method lies in the subjective nature of morphological evaluation. Recent advances in computer vision and artificial intelligence tools offer improved performance and increasing accessibility of these technologies even for diagnostic laboratories.

Method Peripheral blood smears from IPD patients and healthy controls were stained with 13 primary antibodies against platelet structures (granules, cytoskeleton, surface receptors) and 2 fluorescence-labelled secondary antibodies. Immunofluorescence images were taken and processed using a fluorescence microscope (Aklides, Medipan) with bioimage analysis and machine-learning tools such as CLIJ [2], BaSiC [3], ilastik [4] and KNIME [5].

The expression pattern of platelet markers was assessed in IPD patients with respect to controls and eventually compared with the reference reports, which had been previously obtained by traditional manual immunofluorescence analysis.

Results We established a standardized process for the semiautomatic acquisition of immunofluorescence images. The process was able to capture an image stack of 11 images for 6 fields in less than 40 seconds. The image analysis was capable of segmenting platelets and extracting a spectrum of parameters (size, circularity, granularity, fluorescence intensity index). An exposure algorithm for the optimal use of image depth was also applied. Furthermore, artefacts such as bright accumulations of fluorophores were automatically rejected by dividing the image into 16 areas and disregarding the brightest two of them for auto-exposure.

To validate the approach, we blindly assessed 26 healthy controls and 26 patients affected with two IPD characterized by already described immunofluorescence pattern based on the absence or reduced expression of a platelet surface glycoprotein (GP): GP IbIX in Bernard-Soulier syndrome and GP IIbIIIa in Glanzmann thrombasthenia, respectively [1]. The semiautomatic method was able to distinguish all patients (26/26) from controls and to identify the typical diagnostic fluorescence patterns of the disorders.

Conclusion Semiautomatic image analysis allows an unbiased and standardized immunofluorescence based diagnostic process for patients with IPD. Implementation of the method for diagnostic screening routine is becoming a realistic option.

Publikationsverlauf

Artikel online veröffentlicht:
26. Februar 2024

• References

• 1 Zaninetti C, Leinøe E, Lozano ML, Rossing M, Bastida JM, Zetterberg E, Rivera J, Greinacher A.. Validation of immunofluorescence analysis of blood smears in patients with inherited platelet disorders. J Thromb Haemost 2023; 21 (04) 1010-1019
• 2 Haase R, Royer LA, Steinbach P, Schmidt D, Dibrov A, Schmidt U, Weigert M, Maghelli N, Tomancak P, Jug F, Myers EW.. CLIJ: GPU-accelerated image processing for everyone. Nat Methods 2020; 17: 5-6
• 3 Peng T, Thorn K, Schroeder T, Wang L, Theis FJ, Marr C, Navab N.. A BaSiC tool for background and shading correction of optical microscopy images. Nat Commun 2017; 8: 14836
• 4 Berg S, Kutra D, Kroeger T, Straehle CN, Kausler BX, Haubold C, Schiegg M, Ales J, Beier T, Rudy M, Eren K, Cervantes JI, Xu B, Beuttenmueller F, Wolny A, Zhang C, Koethe U, Hamprecht FA, Kreshuk A.. ilastik: interactive machine learning for (bio)image analysis. Nat Methods 2019; 16 (12) 1226-1232
• 5 Berthold MR, Cebron N, Dill F, Gabriel TR, Kötter T, Meinl T, Ohl P, Sieb C, Thiel K, Wiswedel B.. KNIME: The Konstanz Information Miner. In: Preisach C, Burkhardt H, Schmidt-Thieme L, Decker R, eds. Data Analysis, Machine Learning and Applications. Studies in Classification, Data Analysis, and Knowledge Organization. Springer; 2008: 319-326