RESUMEN

BACKGROUND: Convolutional neural network-based image processing research is actively being conducted for pathology image analysis. As a convolutional neural network model requires a large amount of image data for training, active learning (AL) has been developed to produce efficient learning with a small amount of training data. However, existing studies have not specifically considered the characteristics of pathological data collected from the workplace. For various reasons, noisy patches can be selected instead of clean patches during AL, thereby reducing its efficiency. This study proposes an effective AL method for cancer pathology that works robustly on noisy datasets. METHODS: Our proposed method to develop a robust AL approach for noisy histopathology datasets consists of the following three steps: 1) training a loss prediction module, 2) collecting predicted loss values, and 3) sampling data for labeling. This proposed method calculates the amount of information in unlabeled data as predicted loss values and removes noisy data based on predicted loss values to reduce the rate at which noisy data are selected from the unlabeled dataset. We identified a suitable threshold for optimizing the efficiency of AL through sensitivity analysis. RESULTS: We compared the results obtained with the identified threshold with those of existing representative AL methods. In the final iteration, the proposed method achieved a performance of 91.7% on the noisy dataset and 92.4% on the clean dataset, resulting in a performance reduction of less than 1%. Concomitantly, the noise selection ratio averaged only 2.93% on each iteration. CONCLUSIONS: The proposed AL method showed robust performance on datasets containing noisy data by avoiding data selection in predictive loss intervals where noisy data are likely to be distributed. The proposed method contributes to medical image analysis by screening data and producing a robust and effective classification model tailored for cancer pathology image processing in the workplace.

Asunto(s)

RESUMEN

Cervical cancer (CC) is the fourth most common malignant tumor among women worldwide. Constructing a high-accuracy deep convolutional neural network (DCNN) for cervical cancer screening and diagnosis is important for the successful prevention of cervical cancer. In this work, we proposed a robust DCNN for cervical cancer screening using whole-slide images (WSI) of ThinPrep cytologic test (TCT) slides from 211 cervical cancer and 189 normal patients. We used an active learning strategy to improve the efficiency and accuracy of image labeling. The sensitivity, specificity, and accuracy of the best model were 96.21%, 98.95%, and 97.5% for CC patient identification respectively. Our results also demonstrated that the active learning strategy was superior to the traditional supervised learning strategy in cost reduction and improvement of image labeling quality. The related data and source code are freely available at https://github.com/hqyone/cancer_rcnn.

RESUMEN

BACKGROUND: Interpretive pedagogy with simulation encourages students to consider multiple perspectives contextually leading students to think deeper in a shared learning environment. PROBLEM: Clinical sites were lacking in a senior nursing leadership and management course and necessitated the adaptation of traditional clinical teaching methodologies. APPROACH: Low-fidelity simulation was used as an active learning strategy to fulfill clinical hours. OUTCOMES: Comparing student groups' pretest mean scores were not significant (p = .610; 95% confidence interval [CI] [-0.95, 0.12]). Comparatively, the student groups' posttest scores ranging between 87% and 90%, respectively, were also not statistical significance (p = .136, 95% CI [-0.95, 0.12]). CONCLUSION: Students were positive about their experience. They appreciated the opportunity to practice what they learned in the classroom in a safe environment. As a result, simulation in a senior nursing leadership course can be successfully used as an alternative to traditional clinical experiences and fulfill clinical hour requirements.

Asunto(s)

RESUMEN

OBJECTIVE: To test the hypothesis that small group concept mapping of 4 core neonatal topics as part of a fourth-year allopathic medical student elective would improve critical thinking (CT) as measured by the California Critical Thinking Skills Test (CCTST). To describe any correlations between scores on the CCTST and the step 1 and step 2 Clinical Knowledge parts of the United States Medical Licensing Exam. METHODS: Twenty-seven students participated in this pilot study during a 1-month elective. A pretest CCTST, California Critical Thinking Disposition Inventory (CCTDI), and multiple choice knowledge test (MCKT) were completed immediately before the elective began. Four weekly group sessions were held with assigned reading on each of the 4 neonatal topics. Concept mapping was performed in small groups of 4 to 6 students with a group concept map collected at the end of the exercise. A posttest CCTST and MCKT was completed after the 4 group sessions. RESULTS: Pre-CCTST overall score was 83.9 ± 6, and post-CCTST overall score was 85.6 ± 6.9 (P = .57). Pearson correlation of USMLE step 1 and pre-CCTST showed r(25) = .276, P = .164. Pearson correlation of USMLE step 2 CK and pre-CCTST revealed r(25) = .214, P = .482. The precourse MCKT average was 35%, and the postcourse average 50% (P ≤ .001). CONCLUSIONS: A recent meta-analysis confirms this is the first report of a comparison between the increasingly common CCTST and the USMLE. We confirmed that concept mapping is a valid mechanism to teach content knowledge. Although the difference in the CCTST scores was not significant, this study could serve as an important start toward development of a curriculum devoted to teaching content and improving CT. The small number of students may have prevented us from defining a significant impact.

Asunto(s)