Lung cancer has brought tough challenges to human health due to its high incidence and mortality rate in the current practice. Nowadays, computed tomography (CT) imaging is still the most preferred diagnostic tool for early screening of lung cancer. However, a great challenge brought from accumulative CT imaging data can not meet the demand of the current clinical practice. As a novel kind of artificial intelligence technique aimed to deal with medical images, a computer-aided diagnosis has been found to provide useful auxiliary information, attenuate the workload of doctors, and significantly improve the efficiency and accuracy for clinical diagnosis of lung cancer. Therefore, an effective combination of computer-aided techniques and CT imaging has increasingly become an active area of investigation in early diagnosis of lung cancer. This review aims to summarize the latest progress on the diagnostic value of computer-aided technology with regard to early stage lung cancer from the perspectives of machine learning and deep learning.
Clinically, non-contrastive computed tomography (NCCT) is used to quickly diagnose the type and area of stroke, and the Alberta stroke program early computer tomography score (ASPECTS) is used to guide the next treatment. However, in the early stage of acute ischemic stroke (AIS), it’s difficult to distinguish the mild cerebral infarction on NCCT with the naked eye, and there is no obvious boundary between brain regions, which makes clinical ASPECTS difficult to conduct. The method based on machine learning and deep learning can help physicians quickly and accurately identify cerebral infarction areas, segment brain areas, and operate ASPECTS quantitative scoring, which is of great significance for improving the inconsistency in clinical ASPECTS. This article describes current challenges in the field of AIS ASPECTS, and then summarizes the application of computer-aided technology in ASPECTS from two aspects including machine learning and deep learning. Finally, this article summarizes and prospects the research direction of AIS-assisted assessment, and proposes that the computer-aided system based on multi-modal images is of great value to improve the comprehensiveness and accuracy of AIS assessment, which has the potential to open up a new research field for AIS-assisted assessment.
Pectus carinatum (PC) is one of the most common chest wall anomalies, which is characterized by the protrusion of the anterior chest wall including the sternum and adjacent costal cartilages. Mildly patients suffer from mental problems such as self-abasement, while severely suffering patients are disturbed by significant cardiopulmonary symptoms. The traditional Haller index, which is widely used clinically to evaluate the severity of PC, is deficient in diagnosis efficiency and classification. This paper presents an improved Haller index algorithm for PC: first, the contour of the patient chest in the axial computed tomography (CT) slice where the most convex thorax presents is extracted; and then a cubic B-spline curve is employed to fit the extracted contour followed by an eclipse fitting procedure; finally, the improved Haller index and the classification index are automatically calculated based on the analytic curves. The results of CT data analysis using 22 preoperative and postoperative patient CT datasets show that the proposed diagnostic index for PC can diagnose and classify PC patients correctly, which confirms the feasibility of the evaluation index. Furthermore, digital measurement techniques can be employed to improve the diagnostic efficiency of PC, achieving one small step towards the computer-aided intelligent diagnosis and treatment for pediatric chest wall malformations.
Otitis media is one of the common ear diseases, and its accurate diagnosis can prevent the deterioration of conductive hearing loss and avoid the overuse of antibiotics. At present, the diagnosis of otitis media mainly relies on the doctor's visual inspection based on the images fed back by the otoscope equipment. Due to the quality of otoscope equipment pictures and the doctor's diagnosis experience, this subjective examination has a relatively high rate of misdiagnosis. In response to this problem, this paper proposes the use of faster region convolutional neural networks to analyze clinically collected digital otoscope pictures. First, through image data enhancement and preprocessing, the number of samples in the clinical otoscope dataset was expanded. Then, according to the characteristics of the otoscope picture, the convolutional neural network was selected for feature extraction, and the feature pyramid network was added for multi-scale feature extraction to enhance the detection ability. Finally, a faster region convolutional neural network with anchor size optimization and hyperparameter adjustment was used for identification, and the effectiveness of the method was tested through a randomly selected test set. The results showed that the overall recognition accuracy of otoscope pictures in the test samples reached 91.43%. The above studies show that the proposed method effectively improves the accuracy of otoscope picture classification, and is expected to assist clinical diagnosis.
The possibility of solitary pulmonary nodules tending to lung cancer is very high in the middle and late stage. In order to detect the middle and late solitary pulmonary nodules, we present a new computer-aided diagnosis method based on the geometric features. The new algorithm can overcome the disadvantage of the traditional algorithm which can't eliminate the interference of vascular cross section. The proposed algorithm was implemented by multiple clustering of the extracted geometric features of region of interest (ROI) through K-means algorithm, including degree of slenderness, similar degree of circle, degree of compactness and discrete degree. The 232 lung CT images were selected from Lung Image Database Consortium (LIDC) database to do contrast experiment. Compared with the traditional algorithm, the detection rate of the new algorithm was 92.3%, and the error rate was 14.8%. At the same time, the detection rate of the traditional algorithm was only 83.9%, and the error rate was 78.2%. The results show that the proposed algorithm can mark the solitary pulmonary nodules more accurately and reduce the error rate due to precluding the disturbance of vessel section.
To enhance the accuracy of computer-aided diagnosis of adolescent depression based on electroencephalogram signals, this study collected signals of 32 female adolescents (16 depressed and 16 healthy, age: 16.3 ± 1.3) with eyes colsed for 4 min in a resting state. First, based on the phase synchronization between the signals, the phase-locked value (PLV) method was used to calculate brain functional connectivity in the θ and α frequency bands, respectively. Then based on the graph theory method, the network parameters, such as strength of the weighted network, average characteristic path length, and average clustering coefficient, were calculated separately (P < 0.05). Next, using the relationship between multiple thresholds and network parameters, the area under the curve (AUC) of each network parameter was extracted as new features (P < 0.05). Finally, support vector machine (SVM) was used to classify the two groups with the network parameters and their AUC as features. The study results show that with strength, average characteristic path length, and average clustering coefficient as features, the classification accuracy in the θ band is increased from 69% to 71%, 66% to 77%, and 50% to 68%, respectively. In the α band, the accuracy is increased from 72% to 79%, 69% to 82%, and 65% to 75%, respectively. And from overall view, when AUC of network parameters was used as a feature in the α band, the classification accuracy is improved compared to the network parameter feature. In the θ band, only the AUC of average clustering coefficient was applied to classification, and the accuracy is improved by 17.6%. The study proved that based on graph theory, the method of feature optimization of brain function network could provide some theoretical support for the computer-aided diagnosis of adolescent depression.
Automatic classification of different types of cough plays an important role in clinical. In the previous research of cough classification or cough recognition, traditional Mel frequency cepstrum coefficients (MFCC) which extracts feature mainly from low frequency band is usually used as feature expression. In this paper, by analyzing the distributions of spectral energy of dry/wet cough, it is found that spectral difference of two types of cough exits mainly in middle frequency band and high frequency band. To better reflect the spectral difference of dry cough and wet cough, an improved method of extracting reverse MFCC is proposed. In this method, reverse Mel filter-bank in which filters are allocated in reverse Mel scale is adopted and is improved by placing filters only in the frequency band with high spectral energy. As a result, features are mainly extracted from the frequency band where two types of cough show both high spectral energy and distinguished difference. Detailed process of accessing improved reverse MFCC was introduced and hidden Markov models trained by 60 dry cough and 60 wet cough were used as cough classification model. Classification experiment results for 120 dry cough and 85 wet cough showed that, compared to traditional MFCC, better classification performance was achieved by the proposed method and the total classification accuracy was raised from 89.76% to 93.66%.
Transrectal contrast-enhanced ultrasound (CEUS) is an important examination for rectal tumors. The inhomogeneity of the CEUS images has important clinical significance. However, there is no objective method to evaluate this index. In this study, a method based on gray-level co-occurrence matrix (GLCM) is proposed to extract texture features of images and grade these images according the inhomogeneity. Specific processes include compressing the gray level of the image, calculating the texture statistics of gray level co-occurrence matrix, combining feature selection and principal component analysis (PCA) for dimensionality reduction, and training and validating quadratic discriminant analysis (QDA). After ten cross-validation, the overall accuracy rate of machine classification was 87.01%, and the accuracy of each level was as follows: Grade Ⅰ 52.94%, Grade Ⅱ 96.48% and Grade Ⅲ 92.35% respectively. The proposed method has high accuracy in judging grade Ⅱ and Ⅲ images, which can help to identify the grade of inhomogeneity of contrast-enhanced ultrasound images of rectal tumors, and may be used to assist clinical doctors in judging the grade of inhomogeneity of contrast-enhanced ultrasound of rectal tumors.
The convolutional neural network (CNN) could be used on computer-aided diagnosis of lung tumor with positron emission tomography (PET)/computed tomography (CT), which can provide accurate quantitative analysis to compensate for visual inertia and defects in gray-scale sensitivity, and help doctors diagnose accurately. Firstly, parameter migration method is used to build three CNNs (CT-CNN, PET-CNN, and PET/CT-CNN) for lung tumor recognition in CT, PET, and PET/CT image, respectively. Then, we aimed at CT-CNN to obtain the appropriate model parameters for CNN training through analysis the influence of model parameters such as epochs, batchsize and image scale on recognition rate and training time. Finally, three single CNNs are used to construct ensemble CNN, and then lung tumor PET/CT recognition was completed through relative majority vote method and the performance between ensemble CNN and single CNN was compared. The experiment results show that the ensemble CNN is better than single CNN on computer-aided diagnosis of lung tumor.
To realize the accurate positioning and quantitative volume measurement of tumor in head and neck tumor CT images, we proposed a level set method based on augmented gradient. With the introduction of gradient information in the edge indicator function, our proposed level set model is adaptive to different intensity variation, and achieves accurate tumor segmentation. The segmentation result has been used to calculate tumor volume. In large volume tumor segmentation, the proposed level set method can reduce manual intervention and enhance the segmentation accuracy. Tumor volume calculation results are close to the gold standard. From the experiment results, the augmented gradient based level set method has achieved accurate head and neck tumor segmentation. It can provide useful information to computer aided diagnosis.