The promise of quantitative phase imaging and machine learning in medical diagnostics: a review

Introduction

Quantitative phase imaging (QPI) is a method of phase-contrast microscopy which quantifies the phase shift that occurs when light passes through an optically dense object (1,2). Unlike other conventional phase-contrast methods, QPI creates a phase shift image independent of the bright field image. The method is very useful in providing quantitative information about cells such as the morphology and cell dry mass (3). This method can, therefore, provide enormous information about cells. In the past, however, applying information obtained from QPI based cell profiling into practical translational solutions has been challenging due to limited access to analytical tools capable of making full sense of this data (4). Recent advances in Artificial intelligence and machine learning (5) suggest opportunities in applying QPI to medical diagnostics (6). The goal of this paper is to discuss artificial intelligence (AI), machine learning, QPI and how these techniques can be combined in medical diagnostics.

AI

AI can be described as machines (or computers) that mimic “cognitive” functions usually associated with the human mind, for example; “learning” and “problem-solving” (7). A more comprehensive definition of AI is “a system’s ability to correctly interpret external data, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation.” (8). It is very common for tasks previously thought to be unachievable by machines to be removed from the definition of AI once machines achieve that task. This is known as the AI effect (9). AI involve the use of algorithms.

Machine learning

Machine Learning, seen as a sub-set of artificial intelligence (10), relies on patterns and inference to study algorithms and statistical models with the goal of performing tasks without explicit instructions (11). In machine learning, the algorithm basically learns the associations of predictive data using examples in data (10). Using exploratory data analysis through unsupervised learning is known as data mining (12).

Deep learning

Deep learning, a machine learning method based on artificial neural networks, involves feeding an algorithm with large data sets with the goal of discovering the representations needed for classification or detection (13). It may be supervised or unsupervised.

Supervised learning

Supervised Learning, is the machine learning task of mapping input to an output based on input-output pairs (7). The outputs of interest in supervised learning are typically defined by a human supervisor. The goal of learning associations is to be able to predict future outputs based on input after the algorithm has been trained to make predictions.

Unsupervised learning

Unsupervised learning, unlike supervised learning, learns associations in a dataset without any external definition of associations of interest. It, therefore, identifies previously undiscovered associations instead of relying on known ones (10).

Quantitative phase imaging and deep learning in cell diagnostics

The diverse properties of cells in the body changes based on pathophysiological conditions (14). It is a known fact for example that various parasitic infections and diseases alter red blood cell properties (15). Alterations in the deformability of red blood cells have for example been documented in cases of malaria, sickle cell anemia, diabetes, just to name a few (16). Complete blood count, the universally accepted method for hematological examination of red cell properties, is population-based and fails to focus on individual red cell properties (4). Profiling individual red cells can, therefore, be a game-changer in screening diseases based on individual red blood cells (17). QPI, a method that uses the differences in intracellular refractive index distribution as an imaging contrast for label-free imaging (18), can serve as a viable method for profiling individual cells. The challenge, however, is that the data will be accompanied by variations in each cell, a challenge which hitherto would mean difficulty in analyzing the resultant data due to lack of proper analytics tools. With the improvement in computing power and deep learning however, we can increasingly make sense of this massive data. For example, QPI and machine learning have been combined to help in the diagnosis of malaria infections (19), sickle cell (20), and sperm cell analysis (21). It is therefore proposed that, in the future, QPI can be combined with machine learning in order to provide novel diagnostic solutions.

Ethical considerations

It is obvious that large amounts of data sets will be needed in order to these train machine learning algorithms. It is essential that biases in the patient selection are avoided in order to ensure that samples used are representative enough. It is also important to ensure that data used in AI research are ethically sourced and not weaponized to attack vulnerable populations.

Final thoughts

There is great promise in combining QPI with machine learning and it can potentially lead to affordable, groundbreaking diagnostic solutions.

Acknowledgments

Funding: None.

Footnote

Conflicts of Interest: The author has completed the ICMJE uniform disclosure form (available at http://dx.doi.org/10.21037/jmai.2019.10.05). The author has no conflicts of interest to declare.

Ethical Statement: The author is accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Rubin SE. Plant Microtechnique and Microscopy. Oxford University Press. 1999.
2. Cuche E, Bevilacqua F, Depeursinge C. Digital holography for quantitative phase-contrast imaging. Opt Lett 1999;24:291-3. [Crossref] [PubMed]
3. Zangle T, Teitell M. Live-cell mass profiling: an emerging approach in quantitative biophysics. Nat Methods 2014;11:1221-8. [Crossref] [PubMed]
4. Kim G, Jo Y, Cho H, et al. Learning-based screening of hematologic disorders using quantitative phase imaging of individual red blood cells. Biosens Bioelectron 2019;123:69-76. [Crossref] [PubMed]
5. Bishop CM. Pattern Recognition and Machine Learning. Springer. 2006.
6. Cruz JA, Wishart DS. Applications of machine learning in cancer prediction and prognosis. Cancer Inform 2007;2:59-77. [PubMed]
7. Russell SJ, Norvig P. Artificial Intelligence: A Modern Approach. 3rd Edition. Upper Saddle River, New Jersey: Prentice Hall. 2009.
8. Kaplan Andreas, Haenlein Michael. Siri, Siri, in my hand: Who's the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Business Horizons 2009;62:15-25. [Crossref]
9. McCorduck, Pamela. Machines Who Think. 2nd edition. Natick, MA: A. K. Peters, Ltd. 2004.
10. Panch T, Szolovits P, Atun R. Artificial intelligence, machine learning and health systems. J Glob Health 2018;8:020303. [Crossref] [PubMed]
11. Koza JR, Bennett FH, Andre D, et al. Automated Design of Both the Topology and Sizing of Analog Electrical Circuits Using Genetic Programming. In: Gero JS, Sudweeks F. (eds) Artificial Intelligence in Design ’96. Springer, Dordrecht 1996.
12. Friedman JH. Data Mining and Statistics: What's the connection?. Available online: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.374.7489&rep=rep1&type=pdf
13. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015;521:436-44. [Crossref] [PubMed]
14. Diez-Silva M, Dao M, Han J, et al. Shape and Biomechanical Characteristics of Human Red Blood Cells in Health and Disease. MRS Bull 2010;35:382-8. [Crossref] [PubMed]
15. Chandramohanadas R, Park Y, Lui L, et al. Biophysics of malarial parasite exit from infected erythrocytes. PLoS One 2011;6:e20869. [Crossref] [PubMed]
16. Tomaiuolo G. Biomechanical properties of red blood cells in health and disease towards microfluidics. Biomicrofluidics 2014;8:051501. [Crossref] [PubMed]
17. Weatherall DJ. Systems Biology and Red Cells. New Engl J Med 2011;364:376-7. [Crossref] [PubMed]
18. Wolf E. Three-dimensional structure determination of semi-transparent objects from holographic data. Opt Commun 1969;1:153-6. [Crossref]
19. Go T, Kim JH, Byeon H, et al. Machine learning-based in-line holographic sensing of unstained malaria- infected red blood cells. J Biophotonics 2018;11:e201800101. [Crossref] [PubMed]
20. Javidi B, Markman A, Rawat S, et al. Sickle cell disease diagnosis based on spatio-temporal cell dynamics analysis using 3D printed shearing digital holographic microscopy. Opt Express 2018;26:13614-27. [Crossref] [PubMed]
21. Mirsky SK, Barnea I, Levi M, et al. Automated analysis of individual sperm cells using stain-free interferometric phase microscopy and machine learning. Cytometry A 2017;91:893-900. [Crossref] [PubMed]