Research on the Auxiliary Classification and Diagnosis of Lung Cancer Subtypes Based on Histopathological Images

Abstract

Lung cancer (LC) is one of the most serious cancers threatening human health. Histopathological examination is the gold standard for qualitative and clinical staging of lung tumors. However, the process for doctors to examine thousands of histopathological images is very cumbersome, especially for doctors with less experience. Therefore, objective pathological diagnosis results can effectively help doctors choose the most appropriate treatment mode, thereby improving the survival rate of patients. For the current problem of incomplete experimental subjects in the computer-aided diagnosis of lung cancer subtypes, this study included relatively rare lung adenosquamous carcinoma (ASC) samples for the first time, and proposed a computer-aided diagnosis method based on histopathological images of ASC, lung squamous cell carcinoma (LUSC) and small cell lung carcinoma (SCLC). Firstly, the multidimensional features of 121 LC histopathological images were extracted, and then the relevant features (Relief) algorithm was used for feature selection. The support vector machines (SVMs) classifier was used to classify LC subtypes, and the receiver operating characteristic (ROC) curve and area under the curve (AUC) were used to make it more intuitive evaluate the generalization ability of the classifier. Finally, through a horizontal comparison with a variety of mainstream classification models, experiments show that the classification effect achieved by the Relief-SVM model is the best. The LUSC-ASC classification accuracy was 73.91%, the LUSC-SCLC classification accuracy was 83.91% and the ASC-SCLC classification accuracy was 73.67%. Our experimental results verify the potential of the auxiliary diagnosis model constructed by machine learning (ML) in the diagnosis of LC.

AI-based Carcinoma Detection and Classification Using Histopathological Images: A Systematic Review

“…Hence, we could observe a variation in model performance [76]. In a few studies [45,46], the model performance was less as compared to other lung cancer classification models because of microscopic images from private hospitals with small sized datasets. The studies [10,41,44] reported the use of only cell-level information from the images captured at 40X magnification for classification, which resulted in model accuracy of around 85%.…”

“…), autoencoders, and customized CNNs top the list. SVM [45,46,67,86,88,95,96,99,108] and RF tree classifier [45,76,80,82,84,86,103,108] were second and third in the list. In the 'Others' category, we included hashing, LR, quasi-supervised learning algorithm, ensemble learning, and linear classifier based on different features extracted from ROI for decision-making.…”

“…Classification of lung cancer subtypes as ADC, SCC, large cell carcinoma, and small cell carcinoma was based on morphological features of tumor tissue, particular tumor cells, and surrounding tissue environment [10,42]. Several researchers [41][42][43][44][45][46] have reported the use of traditional ML approaches such as SVM [41,46], RF [45], AdaBoost, cw-Boost [43] fast adaptive neural classifier [44], and statistical model [42] to distinguish ADC and SCC. The classification was performed based on various features extracted either from cells, nuclei, or images.…”

“…The features such as texture features, morphological features, Hu's moments, GLGCM, wavelet transform, GLCM, LBP, GLDS, Markov random field, VGGNet features, HOG, GIST, LeNet features, and densitometric features were reported in the literature [10, 41-44, 46, 47]. Li et al [46] applied Lasso and Relief feature selection methods and LDA feature reduction technique on the set of features extracted from tissue samples. In [45], Nishio et al classified the lung cancer subtypes using the homology-based feature extraction method.…”

“…Segmentation of nuclei using Otsu's thresholding method was reported [41,44]. Some research groups proposed feature enhancement methods to improve classification accuracy [43,44,46].…”