PseudoCell: Hard Negative Mining as Pseudo Labeling for Deep Learning-Based Centroblast Cell Detection

Patch classification models based on deep learning have been utilized in whole-slide images (WSI) of H&E-stained tissue samples to assist pathologists in grading follicular lymphoma patients. However, these approaches still require pathologists to manually identify centroblast cells and provide refined labels for optimal performance. To address this, we propose PseudoCell, an object detection framework to automate centroblast detection in WSI (source code is available at https://github.com/IoBT-VISTEC/PseudoCell.git). This framework incorporates centroblast labels from pathologists and combines them with pseudo-negative labels obtained from undersampled false-positive predictions using the cell's morphological features. By employing PseudoCell, pathologists' workload can be reduced as it accurately narrows down the areas requiring their attention during examining tissue. Depending on the confidence threshold, PseudoCell can eliminate 58.18-99.35% of non-centroblasts tissue areas on WSI. This study presents a practical centroblast prescreening method that does not require pathologists' refined labels for improvement. Detailed guidance on the practical implementation of PseudoCell is provided in the discussion section.

ApSense: Data-driven Algorithm in PPG-based Sleep Apnea Sensing

In this paper, we utilized obstructive sleep apnea and cardiovascular disease-related photoplethysmography (PPG) features in constructing the input to deep learning (DL). The features are pulse wave amplitude (PWA), beat-to-beat or RR interval, a derivative of PWA, a derivative of RR interval, systolic phase duration, diastolic phase duration, and pulse area. Then, we develop DL architectures to evaluate the proposed features' usefulness. Eventually, we demonstrate that in human-machine settings where the medical staff only needs to label 20% of the PPG recording length, our proposed features with the developed DL architectures achieve 79.95% and 73.81% recognition accuracy in MESA and HeartBEAT datasets. This simplifies the labelling task of the medical staff during the sleep test yet provides accurate apnea event recognition.