Automated classification of multiphoton microscopy images of ovarian tissue using deep learning

Abstract

Histopathological image analysis of stained tissue slides is routinely used in tumor detection and classification. However, diagnosis requires a highly trained pathologist and can thus be time-consuming, labor-intensive, and potentially risk bias. Here, we demonstrate a potential complementary approach for diagnosis. We show that multiphoton microscopy images from unstained, reproductive tissues can be robustly classified using deep learning techniques. We fine-train four pretrained convolutional neural networks using over 200 murine tissue images based on combined second-harmonic generation and two-photon excitation fluorescence contrast, to classify the tissues either as healthy or associated with high-grade serous carcinoma with over 95% sensitivity and 97% specificity. Our approach shows promise for applications involving automated disease diagnosis. It could also be readily applied to other tissues, diseases, and related classification problems.

Automated classification of hepatocellular carcinoma differentiation using multiphoton microscopy and deep learning

“…In early research, CNNs were mainly applied to white‐light images . In recent years, CNN methods have achieved many breakthroughs in the field of biomedical image processing ; for example, in classification tasks for different diseases , including the automatic diagnosis of lung cancer . Although a variety of CNNs have emerged recently, their basic structure consists of an input layer, convolution layer, activation function, pooling layer, and fully connected layer.…”

“…With the development of deep‐learning techniques for high‐accuracy classification that require fewer training images, the application of CNNs to MPM image data sets has become more viable, and could lead to rapid and reliable automated diagnostic tools. Through a pre‐training network, Huttunen et al and Teikari et al realized the application of CNN to MPM image analysis using a small sample size. Our experiment achieved the classification of high, medium and low differentiation grades of liver cancer in a small sample, with an accuracy approaching 90%.…”

“…In addition, training with larger data sets and more clinical validation (eg, age distribution, gender distribution and so on) will help determine whether other markers play a key role in liver cancer pathologies. Alternatively, other pre‐trained networks can be selected to improve the accuracy of the CNN . Furthermore, considering the abundance of available imaging pathways, such as third‐harmonic generation, coherent anti‐Stokes Raman scattering , and other MPM techniques, there are many other potential routes to obtain label‐free imaging and automated classification of HCC samples.…”

“…Multiphoton microscopy (MPM) has emerged as a promising, rapid, label‐free and high‐resolution microscopic imaging technique, suitable for examining unprocessed tissue samples. It is predicted to be developing into an irreplaceable imaging technology in cancer detection and research . While the technique enables tomographic deep‐imaging of substances in biological tissues using longer‐wavelength infrared lasers, samples can also be observed for longer durations with these lasers because of reduced photodamage.…”

Investigating and Assessing the Dermoepidermal Junction with Multiphoton Microscopy and Deep Learning

“…The employed DL image classification method was inspired from a recent work (Huttunen et al, 2018), and dealt with a total of 358 MPM images from healthy (n = 14) and dysplastic (n = 14) unstained tissue sections, collected to contain the DEJ (see Methods). Images were randomly divided into validation (70%) and training sets (30%), the latter being augmented with the two strategies: (i) by reflecting the original images horizontally and vertically, and (ii) by repeatably blurring these horizontal and vertical reflections by using a five-layer Gaussian image pyramid, (see Methods).…”