Lung cancer lesion detection in histopathology images using graph-based sparse PCA network

Lung cancer detection and classification using optimized CNN features and Squeeze-Inception-ResNeXt model

Lung cancer, with its high mortality rate, is one of the deadliest diseases globally. The alarming increase in lung cancer deaths and its widespread prevalence have led to the development of various cancer control research and early detection methods aimed at reducing mortality rates. Effective diagnostic techniques are crucial for lowering lung cancer incidence, as early detection significantly impacts treatment success. Human error can often impede accurate identification of lung nodules, in which Computer-Aided Diagnostic (CAD) systems are utilized. These systems help radiologists by automating diagnostic processes and improving accuracy of detecting and classifying malignancies. This paper aims to develop a deep learning approach for classifying lung diseases using chest Computed Tomography (CT) scan images. The approach starts with image pre-processing, including color space conversion, data augmentation, resizing, and normalization. Feature extraction is carried out using a Convolutional Neural Network (CNN) optimized with Slime Mould Algorithm (SMA). For classification, a novel approach combining Squeeze-Inception V3 with ResNeXt, referred to as Squeeze-Inception-ResNeXt, is proposed. The Squeeze-Inception-ResNeXt model benefits from reduced computational cost while maintaining high performance in classifying lung diseases. This model categorizes lung diseases into Adenocarcinoma, Large Cell Carcinoma, and Squamous Cell Carcinoma. Additionally, SMA is utilized in training the Squeeze-Inception-ResNeXt model. Experimental results show that Squeeze-Inception-ResNeXt surpasses traditional models, with an accuracy of 97.7 %, sensitivity of 98.1 %, and specificity of 97.4 %.

Four functional genotoxic marker genes (Bax, Btg2, Ccng1, and Cdkn1a) discriminate genotoxic hepatocarcinogens from non-genotoxic hepatocarcinogens and non-genotoxic non-hepatocarcinogens in rat public toxicogenomics data, Open TG-GATEs

BACKGROUND

Previously, Japanese Environmental Mutagen and Genome Society/Mammalian Mutagenicity Study Group/Toxicogenomics Study Group (JEMS/MMS toxicogenomic study group) proposed 12 genotoxic marker genes (Aen, Bax, Btg2, Ccnf, Ccng1, Cdkn1a, Gdf15, Lrp1, Mbd1, Phlda3, Plk2, and Tubb4b) to discriminate genotoxic hepatocarcinogens (GTHCs) from non-genotoxic hepatocarcinogens (NGTHCs) and non-genotoxic non-hepatocarcinogens (NGTNHCs) in mouse and rat liver using qPCR and RNA-Seq and confirmed in public rat toxicogenomics data, Open TG-GATEs, by principal component analysis (PCA). On the other hand, the U.S. Environmental Protection Agency (US EPA) suggested seven genotoxic marker genes (Bax, Btg2, Ccng1, Cgrrf1, Cdkn1a, Mgmt, and Tmem47) with Open TG-GATEs data. Four genes (Bax, Btg2, Ccng1, and Cdkn1a) were common in these two studies. In the present study, we examined the performance of these four genes in Open TG-GATEs data using PCA.

RESULTS

The study's findings are of paramount significance, as these four genes proved to be highly effective in distinguishing five typical GTHCs (2-acetylaminofluorene, aflatoxin B1, 2-nitrofluorene, N-nitrosodiethylamine and N-nitrosomorpholine) from seven typical NGTHCs (clofibrate, ethanol, fenofibrate, gemfibrozil, hexachlorobenzene, phenobarbital, and WY-14643) and 11 NGTNHCs (allyl alcohol, aspirin, caffeine, chlorpheniramine, chlorpropamide, dexamethasone, diazepam, indomethacin, phenylbutazone, theophylline, and tolbutamide) by PCA at 24 h after a single administration with 100% accuracy. These four genes also effectively distinguished two typical GTHCs (2-acetylaminofluorene and N-nitrosodiethylamine) from seven NGTHCs and ten NGTNHCs by PCA on 29 days after 28 days-repeated administrations, with a similar or even better performance compared to the previous 12 genes. Furthermore, the study's analysis revealed that the three intermediate GTHC/NGTHCs (methapyrilene, monocrotaline, and thioacetamide, which were negative in the Salmonella test but positive in the in vivo rat liver test) were located in the intermediate region between typical GTHCs and typical NGTHCs by PCA.

CONCLUSIONS

The present results unequivocally demonstrate the availability of four genotoxic marker genes ((Bax, Btg2, Ccng1, and Cdkn1a) and PCA in discriminating GTHCs from NGTHCs and NGTNHCs in Open TG-GATEs. These findings strongly support our recommendation that future rat liver in vivo toxicogenomics tests prioritize these four genotoxic marker genes, as they have proven to be highly effective in discriminating between different types of hepatocarcinogens.

Colon and lung cancer classification from multi-modal images using resilient and efficient neural network architectures

Automatic classification of colon and lung cancer images is crucial for early detection and accurate diagnostics. However, there is room for improvement to enhance accuracy, ensuring better diagnostic precision. This study introduces two novel dense architectures (D1 and D2) and emphasizes their effectiveness in classifying colon and lung cancer from diverse images. It also highlights their resilience, efficiency, and superior performance across multiple datasets. These architectures were tested on various types of datasets, including NCT-CRC-HE-100K (set of 100,000 non-overlapping image patches from hematoxylin and eosin (H&E) stained histological images of human colorectal cancer (CRC) and normal tissue), CRC-VAL-HE-7K (set of 7180 image patches from N = 50 patients with colorectal adenocarcinoma, no overlap with patients in NCT-CRC-HE-100K), LC25000 (Lung and Colon Cancer Histopathological Image), and IQ-OTHNCCD (Iraq-Oncology Teaching Hospital/National Center for Cancer Diseases), showcasing their effectiveness in classifying colon and lung cancers from histopathological and Computed Tomography (CT) scan images. This underscores the multi-modal image classification capability of the proposed models. Moreover, the study addresses imbalanced datasets, particularly in CRC-VAL-HE-7K and IQ-OTHNCCD, with a specific focus on model resilience and robustness. To assess overall performance, the study conducted experiments in different scenarios. The D1 model achieved an impressive 99.80 % accuracy on the NCT-CRC-HE-100K dataset, with a Jaccard Index (J) of 0.8371, a Matthew's Correlation Coefficient (MCC) of 0.9073, a Cohen's Kappa (Kp) of 0.9057, and a Critical Success Index (CSI) of 0.8213. When subjected to 10-fold cross-validation on LC25000, the D1 model averaged (avg) 99.96 % accuracy (avg J, MCC, Kp, and CSI of 0.9993, 0.9987, 0.9853, and 0.9990), surpassing recent reported performances. Furthermore, the ensemble of D1 and D2 reached 93 % accuracy (J, MCC, Kp, and CSI of 0.7556, 0.8839, 0.8796, and 0.7140) on the IQ-OTHNCCD dataset, exceeding recent benchmarks and aligning with other reported results. Efficiency evaluations were conducted in various scenarios. For instance, training on only 10 % of LC25000 resulted in high accuracy rates of 99.19 % (J, MCC, Kp, and CSI of 0.9840, 0.9898, 0.9898, and 0.9837) (D1) and 99.30 % (J, MCC, Kp, and CSI of 0.9863, 0.9913, 0.9913, and 0.9861) (D2). In NCT-CRC-HE-100K, D2 achieved an impressive 99.53 % accuracy (J, MCC, Kp, and CSI of 0.9906, 0.9946, 0.9946, and 0.9906) with training on only 30 % of the dataset and testing on the remaining 70 %. When tested on CRC-VAL-HE-7K, D1 and D2 achieved 95 % accuracy (J, MCC, Kp, and CSI of 0.8845, 0.9455, 0.9452, and 0.8745) and 96 % accuracy (J, MCC, Kp, and CSI of 0.8926, 0.9504, 0.9503, and 0.8798), respectively, outperforming previously reported results and aligning closely with others. Lastly, training D2 on just 10 % of NCT-CRC-HE-100K and testing on CRC-VAL-HE-7K resulted in significant outperformance of InceptionV3, Xception, and DenseNet201 benchmarks, achieving an accuracy rate of 82.98 % (J, MCC, Kp, and CSI of 0.7227, 0.8095, 0.8081, and 0.6671). Finally, using explainable AI algorithms such as Grad-CAM, Grad-CAM++, Score-CAM, and Faster Score-CAM, along with their emphasized versions, we visualized the features from the last layer of DenseNet201 for histopathological as well as CT-scan image samples. The proposed dense models, with their multi-modality, robustness, and efficiency in cancer image classification, hold the promise of significant advancements in medical diagnostics. They have the potential to revolutionize early cancer detection and improve healthcare accessibility worldwide.

Local heterogeneity of normal lung parenchyma and small airways disease are associated with COPD severity and progression

BACKGROUND

Small airways disease (SAD) is a major cause of airflow obstruction in COPD patients and has been identified as a precursor to emphysema. Although the amount of SAD in the lungs can be quantified using our Parametric Response Mapping (PRM) approach, the full breadth of this readout as a measure of emphysema and COPD progression has yet to be explored. We evaluated topological features of PRM-derived normal parenchyma and SAD as surrogates of emphysema and predictors of spirometric decline.

METHODS

PRM metrics of normal lung (PRMNorm) and functional SAD (PRMfSAD) were generated from CT scans collected as part of the COPDGene study (n = 8956). Volume density (V) and Euler-Poincaré Characteristic (χ) image maps, measures of the extent and coalescence of pocket formations (i.e., topologies), respectively, were determined for both PRMNorm and PRMfSAD. Association with COPD severity, emphysema, and spirometric measures were assessed via multivariable regression models. Readouts were evaluated as inputs for predicting FEV1 decline using a machine learning model.

RESULTS

Multivariable cross-sectional analysis of COPD subjects showed that V and χ measures for PRMfSAD and PRMNorm were independently associated with the amount of emphysema. Readouts χfSAD (β of 0.106, p < 0.001) and VfSAD (β of 0.065, p = 0.004) were also independently associated with FEV1% predicted. The machine learning model using PRM topologies as inputs predicted FEV1 decline over five years with an AUC of 0.69.

CONCLUSIONS

We demonstrated that V and χ of fSAD and Norm have independent value when associated with lung function and emphysema. In addition, we demonstrated that these readouts are predictive of spirometric decline when used as inputs in a ML model. Our topological PRM approach using PRMfSAD and PRMNorm may show promise as an early indicator of emphysema onset and COPD progression.