Automatic glaucoma screening using optic nerve head measurements and random forest classifier on fundus images

Recognition of Glaucomatous Fundus Images Using Machine Learning Methods Based on Optic Nerve Head Topographic Features

PRCIS

Machine learning classifiers are an effective approach to detecting glaucomatous fundus images based on optic disc topographic features making it a straightforward and effective approach.

STUDY DESIGN

Retrospective case-control study.

OBJECTIVE

The aim was to compare the effectiveness of clinical discriminant rules and machine learning classifiers in identifying glaucomatous fundus images based on optic disc topographic features.

METHODS

The study used a total of 800 fundus images, half of which were glaucomatous cases and the other half non-glaucomatous cases obtained from an open database and clinical work. The images were randomly divided into training and testing sets with equal numbers of glaucomatous and non-glaucomatous images. An ophthalmologist framed the edge of the optic cup and disc, and the program calculated five features, including the vertical cup-to-disc ratio and the width of the optic rim in four quadrants in pixels, used to create machine learning classifiers. The discriminative ability of these classifiers was compared with clinical discriminant rules.

RESULTS

The machine learning classifiers outperformed clinical discriminant rules, with the extreme gradient boosting method showing the best performance in identifying glaucomatous fundus images. Decision tree analysis revealed that the cup-to-disc ratio was the most important feature for identifying glaucoma fundus images. At the same time, the temporal width of the optic rim was the least important feature.

CONCLUSIONS

Machine learning classifiers are an effective approach to detecting glaucomatous fundus images based on optic disc topographic features and integration with an automated program for framing and calculating the required parameters would make it a straightforward and effective approach.

Automatic detection of abnormal hand gestures in patients with radial, ulnar, or median nerve injury using hand pose estimation

Background

Radial, ulnar, or median nerve injuries are common peripheral nerve injuries. They usually present specific abnormal signs on the hands as evidence for hand surgeons to diagnose. However, without specialized knowledge, it is difficult for primary healthcare providers to recognize the clinical meaning and the potential nerve injuries through the abnormalities, often leading to misdiagnosis. Developing technologies for automatically detecting abnormal hand gestures would assist general medical service practitioners with an early diagnosis and treatment.

Methods

Based on expert experience, we selected three hand gestures with predetermined features and rules as three independent binary classification tasks for abnormal gesture detection. Images from patients with unilateral radial, ulnar, or median nerve injuries and healthy volunteers were obtained using a smartphone. The landmark coordinates were extracted using Google MediaPipe Hands to calculate the features. The receiver operating characteristic curve was employed for feature selection. We compared the performance of rule-based models with logistic regression, support vector machine and of random forest machine learning models by evaluating the accuracy, sensitivity, and specificity.

Results

The study included 1,344 images, twenty-two patients, and thirty-four volunteers. In rule-based models, eight features were finally selected. The accuracy, sensitivity, and specificity were (1) 98.2, 91.7, and 99.0% for radial nerve injury detection; (2) 97.3, 83.3, and 99.0% for ulnar nerve injury detection; and (3) 96.4, 87.5, and 97.1% for median nerve injury detection, respectively. All machine learning models had accuracy above 95% and sensitivity ranging from 37.5 to 100%.

Conclusion

Our study provides a helpful tool for detecting abnormal gestures in radial, ulnar, or median nerve injuries with satisfying accuracy, sensitivity, and specificity. It confirms that hand pose estimation could automatically analyze and detect the abnormalities from images of these patients. It has the potential to be a simple and convenient screening method for primary healthcare and telemedicine application.

Segmentation technique and dynamic ensemble selection to enhance glaucoma severity detection

The severity of glaucoma can be observed by categorising glaucoma diseases into several classes based on a classification process. The two most suitable parameters, cup-to-disc ratio (CDR) and peripapillary atrophy (PPA), which are commonly used to identify glaucoma are utilized in this study to strengthen the classification. First, an active contour snake (ACS) is employed to retrieve both optic disc (OD) and optic cup (OC) values, which are required to calculate the CDR. Moreover, Otsu segmentation and thresholding techniques are used to identify PPA, and the features are then extracted using a grey-level co-occurrence matrix (GLCM). An advanced segmentation technique, combined with an improved classifier called dynamic ensemble selection (DES), is proposed to classify glaucoma. Because DES is generally used to handle an imbalanced dataset, the proposed model is expected to detect glaucoma severity and determine the subsequent treatment accurately. The proposed model obtains a higher mean accuracy (0.96) than the deep learning-based U-Net (0.90) when evaluated using three datasets of 250 retinal fundus images (200 training, 50 testings) based on the 5-fold cross-validation scheme.

Spotlight on the Disc-Damage Likelihood Scale (DDLS)

The disc damage likelihood scale (DDLS) is a tool for classifying glaucomatous structural changes to the optic disc based on the radial width of the neuroretinal rim at its thinnest location, or if no rim is present, the extent of absence of the rim. Unlike cup disc ratio (CDR), the DDLS also considers disc size. Twenty years after its first description, the aim of this review was to critically appraise evidence for the DDLS and evaluate its role in current practice. A literature search by two independent authors identified 33 relevant articles for inclusion. Five studies evaluated reproducibility, 5 diagnostic performance, and 2 studies examined ability to detect progression. Eleven studies evaluated correlation between DDLS and other markers of glaucoma. Despite the widespread availability of imaging devices such as optical coherence tomography (OCT), clinical examination of the optic disc remains an essential component of glaucoma diagnosis and monitoring. The DDLS provides a reliable method for semi-quantitative clinical grading of the optic disc in glaucoma, with higher reproducibility than methods such as CDR.