Infrared (IR) hyperspectral imaging is a powerful approach in the field of materials and life sciences. However, for the extension to modern sub-diffraction nanoimaging it still remains a highly inefficient technique, as it acquires data via inherent sequential schemes. Here, we introduce the mathematical technique of low-rank matrix reconstruction to the sub-diffraction scheme of atomic force microscopy-based infrared spectroscopy (AFM-IR), for efficient hyperspectral IR nanoimaging. To demonstrate its application potential, we chose the trypanosomatid unicellular parasites Leishmania species as a realistic target of biological importance. The mid-IR spectral fingerprint window covering the spectral range from 1300 to 1900 cm−1 was chosen and a distance between the data points of 220 nm was used for nanoimaging of single parasites. The method of k-means cluster analysis was used for extracting the chemically distinct spatial locations. Subsequently, we randomly selected only 10% of an originally gathered data cube of 134 (x) × 50 (y) × 148 (spectral) AFM-IR measurements and completed the full data set by low-rank matrix reconstruction. This approach shows agreement in the cluster regions between full and reconstructed data cubes. Furthermore, we show that the results of the low-rank reconstruction are superior compared to alternative interpolation techniques in terms of error-metrics, cluster quality, and spectral interpretation for various subsampling ratios. We conclude that by using low-rank matrix reconstruction the data acquisition time can be reduced from more than 14 h to 1–2 h. These findings can significantly boost the practical applicability of hyperspectral nanoimaging in both academic and industrial settings involving nano- and bio-materials.
