Intelligent medicine最新文献

Background

We created and validated a computed tomography (CT)-based radiomic model using both clinical factors and the radiomic signature for assessing the strangulation risk of acute intestinal obstruction. This would assist surgeons in accurately predicting intestinal ischemia and strangulation in patients with intestinal obstruction.

Methods

We recruited 289 patients with acute intestinal obstruction admitted in the Affiliated Hospital of Qingdao University from January 2019 to February 2022. The patients were allocated to a training (n = 226) and validation cohort (n = 63). Radiomic features were collected from CT images, and the radiomic signature was extracted and used to calculate a radiomic score (Rad-score). A nomogram was constructed using the clinical features and the Rad-score, and the performance of the clinical, radiomics, and nomogram models was assessed in the two cohorts.

Results

Six robust features were used to construct the radiomic signature. The nomogram incorporating hemoglobin levels, C-reactive protein levels, American Society of Anesthesiologists score, time of obstruction, CT image of mesenteric fluid (P < 0.05), and the signature demonstrated good predictive ability for intestinal ischemia in patients with acute intestinal obstruction, with areas under the curve of 0.892 (95% confidence interval, 0.837–0.947) and 0.781 (95% confidence interval, 0.619–0.944) for the training and validation sets, respectively. The decision curve analysis showed that this model outperformed the clinical and radiomic signature models.

Conclusion

The radiomic nomogram may effectively predict intestinal ischemia in patients with acute intestinal disease and may assist clinical decision-making.

Objective Diabetes mellitus is a serious disease where the body of affected patients are failed to produce enough insulin that causes an abnormality of blood sugar. This disease happens for a number of reasons including modern lifestyle, lethargic attitude, unhealthy food consumption, family history, age, overweight, etc. The aim of this study was to propose a machine learning based prediction model that detected diabetes at the beginning.

Methods In this work, we collected 520 patients records from the University of California, Irvine (UCI) machine learning repository of Sylhet Diabetes Hospital, Sylhet. Then, a similar questionnaire of that hospital was followed and assembled 558 patients records from all over Bangladesh through this questionnaire. However, we accumulated patient records of these two datasets. In the next step, these datasets were cleaned and applied thirty five state-of-arts classifiers such as logistic regression (LR), K nearest neighbors (KNN), support vector classifier (SVC), Nave Byes (NB), decision tree (DT), random forest (RF), stochastic gradient descent (SGD), Perceptron, AdaBoost, XGBoost, passive aggressive classifier (PAC), ridge classifier (RC), Nu-support vector classifier (Nu-SVC), linear support vector classifier (LSVC), calibrated classifier CV (CCCV), nearest centroid (NC), Gaussian process classifier (GPC), multinomial NB (MNB), complement NB, Bernoulli NB (BNB), categorical NB, Bagging, extra tree(ET), gradiant boosting classifier (GBC), Hist gradiant boosting classifier (HGBC), one vs rest classifier (OVsRC), multi-layer perceptron (MLP), label propagation (LP), label spreading (LS), stacking, ridge classifier CV (RCCV), logistic regression CV (LRCV), linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and light gradient boosting machine (LGBM) to explore best stable predictive model. The performance of the classifiers has been measured using five metrics such as accuracy, precision, recall, F1-score, and area under the receiver operating characteristic. Finally, these outcomes were interpreted using Shapley additive explanations methods and identified relevant features for happening diabetes.

Results In this work, different classifiers were shown their performance where ET outperformed any other classifiers with 97.11% accuracy for the Sylhet Diabetes Hospital dataset (SDHD) and MLP shows the best accuracy (96.42%) for the collected dataset. Subsequently, HGBC and LGBM provide the highest 94.90% accuracy for the combined datasets individually.

Conclusion LGBM, stacking, HGBC, RF, ET, bagging, and GBC might represent more stable prediction results for each dataset.

Objective

Appropriate medical imaging is important for value-based care. We aim to evaluate the performance of generative pretrained transformer 4 (GPT-4), an innovative natural language processing model, providing appropriate medical imaging automatically in different clinical scenarios.

Methods

Institutional Review Boards (IRB) approval was not required due to the use of nonidentifiable data. Instead, we used 112 questions from the American College of Radiology (ACR) Radiology-TEACHES Program as prompts, which is an open-sourced question and answer program to guide appropriate medical imaging. We included 69 free-text case vignettes and 43 simplified cases. For the performance evaluation of GPT-4 and GPT-3.5, we considered the recommendations of ACR guidelines as the gold standard, and then three radiologists analyzed the consistency of the responses from the GPT models with those of the ACR. We set a five-score criterion for the evaluation of the consistency. A paired t-test was applied to assess the statistical significance of the findings.

Results

For the performance of the GPT models in free-text case vignettes, the accuracy of GPT-4 was 92.9%, whereas the accuracy of GPT-3.5 was just 78.3%. GPT-4 can provide more appropriate suggestions to reduce the overutilization of medical imaging than GPT-3.5 (t = 3.429, P = 0.001). For the performance of the GPT models in simplified scenarios, the accuracy of GPT-4 and GPT-3.5 was 66.5% and 60.0%, respectively. The differences were not statistically significant (t = 1.858, P = 0.070). GPT-4 was characterized by longer reaction times (27.1 s in average) and extensive responses (137.1 words on average) than GPT-3.5.

Conclusion

As an advanced tool for improving value-based healthcare in clinics, GPT-4 may guide appropriate medical imaging accurately and efficiently.

Background Oral cancer is one of the most common types of cancer in men causing mortality if not diagnosed early. In recent years, computer-aided diagnosis (CAD) using artificial intelligence techniques, in particular, deep neural networks have been investigated and several approaches have been proposed to deal with the automated detection of various pathologies using digital images. Recent studies indicate that the fusion of images with the patient’s clinical information is important for the final clinical diagnosis. As such dataset does not yet exist for oral cancer, as far as the authors are aware, a new dataset was collected consisting of histopathological images, demographic and clinical data. This study evaluated the importance of complementary data to histopathological image analysis of oral leukoplakia and carcinoma for CAD.

Methods A new dataset (NDB-UFES) was collected from 2011 to 2021 consisting of histopathological images and information. The 237 samples were curated and analyzed by oral pathologists generating the gold standard for classification. State-of-the-art image fusion architectures and complementary data (Concatenation, Mutual Attention, MetaBlock and MetaNet) using the latest deep learning backbones were investigated for 4 distinct tasks to identify oral squamous cell carcinoma, leukoplakia with dysplasia and leukoplakia without dysplasia. We evaluate them using balanced accuracy, precision, recall and area under the ROC curve metrics.

Results Experimental results indicate that the best models present balanced accuracy of $83.24\%$ using images, demographic and clinical information with MetaBlock fusion and ResNetV2 backbone. It represents an improvement in performance of $30.68\%$ (19.54 pp) in the task to differentiate samples diagnosed with oral squamous cell carcinoma and leukoplakia with or without dysplasia.

Conclusion This study indicates that cured demographic and clinical data may positively influence the performance of artificial intelligence models in automated classification of oral cancer.

Objective

To analyze the characteristics of tongue imaging color parameters in patients treated with percutaneous coronary intervention (PCI) and non-PCI for coronary atherosclerotic heart disease (CHD), and to observe the effects of PCI on the tongue images of patients as a basis for the clinical diagnosis and treatment of patients with CHD.

Methods

This study used a retrospective cross-sectional survey to analyze tongue photographs and medical history information from 204 patients with CHD between November 2018 and July 2020. Tongue images of each subject were obtained using the Z-BOX Series traditional Chinese medicine (TCM) intelligent diagnosis instruments, the SMX System 2.0 was used to transform the image data into parameters in the HSV color space, and finally the parameters of the tongue image between patients in the PCI-treated and non-PCI-treated groups for CHD were analyzed.

Results

Among the 204 patients, 112 were in the non-PCI treatment group (38 men and 74 women; average age of (68.76 ± 9.49) years), 92 were in the PCI treatment group (66 men and 26 women; average age of (66.02 ± 10.22) years). In the PCI treatment group, the H values of the middle and tip of the tongue and the overall coating of the tongue were lower (P < 0.05), while the V values of the middle, tip, both sides of the tongue, the whole tongue and the overall coating of the tongue were higher (P < 0.05).

Conclusion

The color parameters of the tongue image could reflect the physical state of patients treated with PCI, which may provide a basis for the clinical diagnosis and treatment of patients with CHD.