Search

Objective: Colorectal cancer (CRC) patients respond differently to treatments and are sub-classified by different approaches. We evaluated a deep learning model, which adopted endoscopic knowledge learnt from AI-doscopist, to characterise CRC patients by histopathological features. Results: Data of 461 patients were collected from TCGA-COAD database. The proposed framework was able to 1) differentiate tumour from normal tissues with an Area Under Receiver Operating Characteristic curve (AUROC) of 0.97; 2) identify certain gene mutations (MYH9, TP53) with an AUROC > 0.75; 3) classify CMS2 and CMS4 better than the other subtypes; and 4) demonstrate the generalizability of predicting KRAS mutants in an external cohort. Conclusions: Artificial intelligent can be used for on-site patient classification. Although KRAS mutants were commonly associated with therapeutic resistance and poor prognosis, subjects with predicted KRAS mutants in this study have a higher survival rate in 30 months after diagnoses.

Blood pressure (BP) is an important clinical vital sign that varies from beat-to-beat. Nevertheless, these variations cannot be captured by the conventional cuff-based BP monitors. This study proposes and evaluates novel cuffless frameworks to continuously estimate the 10-beat averaged systolic BP (SBP) and diastolic BP (DBP) during dynamic exercise by fusing information from multiple biosensors using five machine learning algorithms. Over 100 thousand beats of data were collected from 62 subjects (aged 59 ± 10 years), each underwent a maximal exercise stress test. The average length of recording for each subject was 35 minutes. The BP ranges were 75-280 mmHg for SBP and 36-157 mmHg for DBP respectively. Multiple physiological parameters were measured continuously and used as inputs to five machine learning algorithms for estimating the 10-beat SBP and DBP averages before, during and after the cycling exercise. The mean absolute error (MAE) of Gaussian process regression (GPR) model was 4.8 mmHg and 3.4 mmHg for SBP and DBP, respectively. The MAE of multiple linear regression (MLR), regression tree (RT), ensemble of trees (ETs), and support vector machine (SVM) models varied from 6.1 mmHg to 17.6 mmHg and from 4.0 mmHg to 9.7 mmHg for SBP and DBP, respectively. The GPR model outperformed the other four models and showed promising results in estimating the 10-beat averages of both SBP and DBP without a cuff in a general elderly population under dynamic conditions.

Abstract We have designed a deep-learning model, an “Artificial Intelligent Endoscopist (a.k.a. AI-doscopist)”, to localise colonic neoplasia during colonoscopy. This study aims to evaluate the agreement between endoscopists and AI-doscopist for colorectal neoplasm localisation. AI-doscopist was pre-trained by 1.2 million non-medical images and fine-tuned by 291,090 colonoscopy and non-medical images. The colonoscopy images were obtained from six databases, where the colonoscopy images were classified into 13 categories and the polyps’ locations were marked image-by-image by the smallest bounding boxes. Seven categories of non-medical images, which were believed to share some common features with colorectal polyps, were downloaded from an online search engine. Written informed consent were obtained from 144 patients who underwent colonoscopy and their full colonoscopy videos were prospectively recorded for evaluation. A total of 128 suspicious lesions were resected or biopsied for histological confirmation. When evaluated image-by-image on the 144 full colonoscopies, the specificity of AI-doscopist was 93.3%. AI-doscopist were able to localise 124 out of 128 polyps (polyp-based sensitivity = 96.9%). Furthermore, after reviewing the suspected regions highlighted by AI-doscopist in a 102-patient cohort, an endoscopist has high confidence in recognizing four missed polyps in three patients who were not diagnosed with any lesion during their original colonoscopies. In summary, AI-doscopist can localise 96.9% of the polyps resected by the endoscopists. If AI-doscopist were to be used in real-time, it can potentially assist endoscopists in detecting one more patient with polyp in every 20–33 colonoscopies.

Security of the emerging body sensor network (BSN) in telemedicine applications is a crucial problem because personal medical information must be protected against flaws and misdeeds. The solution is, however, nontrivial because lightweight mechanisms have to be deployed to meet the stringent resource constraints of these networks. It has been suggested that the inherent ability of human body to transfer information is a unique and resource-saving method to secure wireless communications within a BSN. For example, physiological characteristics can be captured by individual sensors of a BSN to generate entity identifiers (EIs) for identifying nodes and even securing keying materials, i.e., by a biometric approach. This study demonstrates the performance analysis of such a biometric trait, i.e., the interpulse intervals (IPIs) of heartbeats that were calculated from electrocardiogram and photoplethysmogram of 99 subjects. Based on the characteristics of IPIs, a lightweight generation scheme of EIs is proposed. Individual randomness and group similarity of the generated EIs are then evaluated. False acceptance rate and false rejection rate are also calculated to measure the effectiveness of the proposed identification system. The results suggest that the readily available IPI information can be a good source for generating EIs among BSN nodes. [ABSTRACT FROM AUTHOR]