Screening chronic myeloid leukemia neutrophils using a novel 3-Dimensional Spectral Gradient Mapping algorithm on hyperspectral images. (June 2022)

• Record Type: Journal Article
• Authors: Panda, Amrit; Pachori, Ram Bilas; Kakkar, Naveen; Joseph John, M; Sinnappah-Kang, Neeta Devi
• Is Part Of: Computer methods and programs in biomedicine. Volume 220(2022)
• Publication Date: 2022-06
• Subjects: Principal component analysis -- 3-D Spectral Gradient Mapping -- Windowed Spectral Angle Mapping -- Hyperspectral image processing -- Chronic myeloid leukemia; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2022.106836
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: Screening of Chronic Myeloid Leukemia using hyperspectral imaging. Novel 3-D Spectral Gradient Mapping algorithm to detect Chronic Myeloid Leukemia. Hyperspectral imaging in minimally invasive diagnosis of blood-based cancer. Disease specific signature using hyperspectral imaging on blood smears. High specificity and accuracy in Chronic Myeloid Leukemia detection using hyperspectral imaging. Abstract: Background and objective Early diagnosis of chronic myeloid leukemia (CML) is important for effective treatment. The high spectral and spatial resolution of hyperspectral cellular or tissue images coupled with image analysis algorithms may provide avenues to detect and diagnose diseases early. Many algorithms have been used to analyze medical hyperspectral image data, each having their own strengths and short-comings. We present a novel 3-Dimensional Spectral Gradient Mapping (3-D SGM) method to analyze hyperspectral image cubes of CML versus healthy blood smears. Methods In the present study, we analyzed 13 hyperspectral image cubes of CML and healthy neutrophils. The 3-D SGM algorithm was compared to the conventional Windowed Spectral Angle Mapping (Windowed SAM) method. The 3-D SGM exploited the spectral information of the image cube together with the inter-band and inter-pixel data by extracting the 3-D gradient vector from each pixel. The Windowed SAM determined the similarity between the averaged window of a 2 × 2 training pixel group and the test pixel, in the multidimensional spectral angle. Results The specificity measure of 3-D SGM (97.7%) was superior to Windowed SAM (72.7%) at ruling out the presence of the disease, making it potentially ideal for screening patients. The positive likelihood ratio value of 3-D SGM (16.70) was superior in diagnosing the presence of the disease (i.e., positive test for CML) versus Windowed SAM (2.26). An accuracy value of 84.2% was achieved with 3-D SGM versus only 70.2% for Windowed SAM. Conclusion The new method is efficient and robust for analyzing hyperspectral images of CML versus healthy neutrophils. It has the potential to be developed into an inexpensive, minimally invasive method for screening CML, and could directly facilitate early diagnosis and treatment of the disease.