Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

Pathology Pathology

Robust Semi-Supervised Learning for Histopathology Images through Self-Supervision Guided Out-of-Distribution Scoring

ArXiv Preprint

Semi-supervised learning (semi-SL) is a promising alternative to supervised learning for medical image analysis when obtaining good quality supervision for medical imaging is difficult. However, semi-SL assumes that the underlying distribution of unaudited data matches that of the few labeled samples, which is often violated in practical settings, particularly in medical images. The presence of out-of-distribution (OOD) samples in the unlabeled training pool of semi-SL is inevitable and can reduce the efficiency of the algorithm. Common preprocessing methods to filter out outlier samples may not be suitable for medical images that involve a wide range of anatomical structures and rare morphologies. In this paper, we propose a novel pipeline for addressing open-set supervised learning challenges in digital histology images. Our pipeline efficiently estimates an OOD score for each unlabelled data point based on self-supervised learning to calibrate the knowledge needed for a subsequent semi-SL framework. The outlier score derived from the OOD detector is used to modulate sample selection for the subsequent semi-SL stage, ensuring that samples conforming to the distribution of the few labeled samples are more frequently exposed to the subsequent semi-SL framework. Our framework is compatible with any semi-SL framework, and we base our experiments on the popular Mixmatch semi-SL framework. We conduct extensive studies on two digital pathology datasets, Kather colorectal histology dataset and a dataset derived from TCGA-BRCA whole slide images, and establish the effectiveness of our method by comparing with popular methods and frameworks in semi-SL algorithms through various experiments.

Nikhil Cherian Kurian, Varsha S, Abhijit Patil, Shashikant Khade, Amit Sethi

2023-03-17

General General

Microprotein Dysregulation in the Serum of Patients with Atrial Fibrillation.

In Journal of proteome research

The incidence rate of atrial fibrillation (AF) has stayed at a high level in recent years. Despite the intensive efforts to study the pathologic changes of AF, the molecular mechanism of disease development remains unclarified. Microproteins are ribosomally translated gene products from small open reading frames (sORFs) and are found to play crucial biological functions, while remain rare attention and indistinct in AF study. In this work, we recruited 65 AF patients and 65 healthy subjects for microproteomic profiling. By differential analysis and cross-validation between independent datasets, a total of 4 microproteins were identified as significantly different, including 3 annotated ones and 1 novel one. Additionally, we established a diagnostic model with either microproteins or global proteins by machine learning methods and found the model with microproteins achieved comparable and excellent performance as that with global proteins. Our results confirmed the abnormal expression of microproteins in AF and may provide new perspectives on the mechanism study of AF.

Zhang Zheng, Tian Tao, Pan Ni, Wang Yi, Peng Mingbo, Zhao Xinbo, Pan Zhenwei, Wan Cuihong

2023-Mar-16

APCO1, LC/MS/MS, atrial fibrillation, cardiac homeostasis, machine learning, microproteins, serum proteomics

General General

Deep-Learning Segmentation of Urinary Stones in Non-Contrast Computed Tomography.

In Journal of endourology

Non-contrast computed tomography (NCCT) relies on labour-intensive examinations of CT slices to identify urolithiasis in the urinary tract, so several algorithms including deep learning method were proposed for segmenting urinary stones. The most problematic point is that many non-stone objects were detected together. The proposed method is an end-to-end deep learning algorithm which could reduce these false positive effectively, and showed high segmentation performance. Automated segmentation is not only could reduce the detection burden, but also help physicians to determine the treatment strategy more precisely measuring the size, volume, stone-to-skin distance, mean stone density, etc. in 3d-space.

Kim Young In, Song Sang Hoon, Park Juhyun, Youn Hye Jung, Kweon Jihoon, Park Hyung Keun

2023-Mar-16

oncology Oncology

Artificial intelligence to empower diagnosis of myelodysplastic syndromes by multiparametric flow cytometry.

In Haematologica ; h5-index 65.0

The diagnosis of myelodysplastic syndromes (MDS) might be challenging and relies on the convergence of cytological, cytogenetic, and molecular arguments. Multiparametric flow cytometry (MFC) helps diagnose MDS, especially when other features are non-contributory, but remains underestimated mostly due to a lack of standardization of cytometers. We present here an innovative model integrating artificial intelligence (AI) with MFC to improve the diagnosis and the classification of MDS. We develop a machine learning model by elasticnet algorithm trained on a cohort of 191 patients and only based on flow cytometry parameters selected by Boruta algorithm, to build a simple but reliable prediction score with 5 parameters. Our MDS prediction score assisted by AI greatly improves the sensitivity of Ogata score while keeping an excellent specificity validated on an external cohort of 89 patients with an AUC = 0.935. This model allows the diagnosis of both high and low risk MDS with 91.8% sensitivity and 92.5% specificity. Interestingly, it highlights a progressive evolution of the score from clonal hematopoiesis of indeterminate potential (CHIP) to highrisk MDS, suggesting a linear evolution between these different stages. By significantly decreasing the overall misclassification of 52% for patients with MDS and of 31.3% for those without MDS (p=0.02), our AI-assisted prediction score outperforms the Ogata score and positions itself as a reliable tool to help diagnose myelodysplastic syndromes.

Clichet Valentin, Lebon Delphine, Chapuis Nicolas, Zhu Jaja, Bardet Valérie, Marolleau Jean-Pierre, Garçon Loïc, Caulier Alexis, Boyer Thomas

2023-Mar-16

General General

Deep-Learning to Map a Benchmark Dataset of Non-amputee Ambulation for Controlling an Open Source Bionic Leg.

In IEEE robotics and automation letters

Powered lower-limb prosthetic devices may be becoming a promising option for amputation patients. Although various methods have been proposed to produce gait trajectories similar to those of non-disabled individuals, implementing these control methods is still challenging. It remains unclear whether these methods provide appropriate, safe, and intuitive locomotion as intended. This paper proposes the direct mapping of the voluntary movement of a residual limb (i.e., thigh) to the desired impedance parameters for amputated limbs (i.e., knee and ankle). The proposed model was learned from the gait trajectories of intact limb individuals from a publicly available biomechanics dataset, and was applied to control the prosthetic leg without post-tuning the network. Thus, the proposed method does not require training time with individuals with amputation nor configuration time for its use, and it provides a closely resembling gait trajectory of the intact limb. For preliminary testing, three able-bodied subjects participated in bypass tests. The proposed model accomplished intuitive and reliable level-ground walking at three different step lengths: self-selected, long-, and short-step lengths. The results indicate that intact benchmark data with different sensor configurations can be directly used to train the model to control prosthetic legs.

Kim Minjae, Hargrove Levi J

2022-Oct

Deep Learning Methods, Prosthetics and Exoskeletons

General General

New Metrics from Polysomnography: Precision Medicine for OSA Interventions.

In Nature and science of sleep

Obstructive sleep apnea (OSA) is a highly preventable disease accompanied by multiple comorbid conditions. Despite the well-established cardiovascular and neurocognitive sequelae with OSA, the optimal metric for assessing the OSA severity and response to therapy remains controversial. Although overnight polysomnography (PSG) is the golden standard for OSA diagnosis, the abundant information is not fully exploited. With the development of deep learning and the era of big data, new metrics derived from PSG have been validated in some OSA consequences and personalized treatment. In this review, these metrics are introduced based on the pathophysiological mechanisms of OSA and new technologies. Emphasis is laid on the advantages and the prognostic value against apnea-hypopnea index. New classification criteria should be established based on these metrics and other clinical characters for precision medicine.

Guo Junwei, Xiao Yi

2023

metrics, obstructive sleep apnea, polysomnography, precision medicine