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

oncology Oncology

Moving Forward Liquid Biopsy in Early Liver Cancer Detection.

In Cancer discovery ; h5-index 105.0

Early cancer detection is an attractive and promising application for liquid biopsy that might revolutionize cancer screenings. In this issue of Cancer Discovery, Foda and colleagues expand the potential utility of a machine learning fragmentome-based model, called DELFI, for detecting liver cancer in high-risk patients. See related article by Foda et al., p. 616 (5).

Rolfo Christian, Russo Alessandro

2023-Mar-01

Surgery Surgery

Robust prediction of nonhome discharge following elective anterior cervical discectomy and fusion using explainable machine learning.

In European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society

PURPOSE : Predict nonhome discharge (NHD) following elective anterior cervical discectomy and fusion (ACDF) using an explainable machine learning model.

METHODS : 2227 patients undergoing elective ACDF from 2008 to 2019 were identified from a single institutional database. A machine learning model was trained on preoperative variables, including demographics, comorbidity indices, and levels fused. The validation technique was repeated stratified K-Fold cross validation with the area under the receiver operating curve (AUROC) statistic as the performance metric. Shapley Additive Explanation (SHAP) values were calculated to provide further explainability regarding the model's decision making.

RESULTS : The preoperative model performed with an AUROC of 0.83 ± 0.05. SHAP scores revealed the most pertinent risk factors to be age, medicare insurance, and American Society of Anesthesiology (ASA) score. Interaction analysis demonstrated that female patients over 65 with greater fusion levels were more likely to undergo NHD. Likewise, ASA demonstrated positive interaction effects with female sex, levels fused and BMI.

CONCLUSION : We validated an explainable machine learning model for the prediction of NHD using common preoperative variables. Adding transparency is a key step towards clinical application because it demonstrates that our model's "thinking" aligns with clinical reasoning. Interactive analysis demonstrated that those of age over 65, female sex, higher ASA score, and greater fusion levels were more predisposed to NHD. Age and ASA score were similar in their predictive ability. Machine learning may be used to predict NHD, and can assist surgeons with patient counseling or early discharge planning.

Geng Eric A, Gal Jonathan S, Kim Jun S, Martini Michael L, Markowitz Jonathan, Neifert Sean N, Tang Justin E, Shah Kush C, White Christopher A, Dominy Calista L, Valliani Aly A, Duey Akiro H, Li Gavin, Zaidat Bashar, Bueno Brian, Caridi John M, Cho Samuel K

2023-Feb-28

Anterior cervical discectomy and fusion, Machine learning, Outcomes prediction

General General

New, fast, and precise method of COVID-19 detection in nasopharyngeal and tracheal aspirate samples combining optical spectroscopy and machine learning.

In Brazilian journal of microbiology : [publication of the Brazilian Society for Microbiology]

Fast, precise, and low-cost diagnostic testing to identify persons infected with SARS-CoV-2 virus is pivotal to control the global pandemic of COVID-19 that began in late 2019. The gold standard method of diagnostic recommended is the RT-qPCR test. However, this method is not universally available, and is time-consuming and requires specialized personnel, as well as sophisticated laboratories. Currently, machine learning is a useful predictive tool for biomedical applications, being able to classify data from diverse nature. Relying on the artificial intelligence learning process, spectroscopic data from nasopharyngeal swab and tracheal aspirate samples can be used to leverage characteristic patterns and nuances in healthy and infected body fluids, which allows to identify infection regardless of symptoms or any other clinical or laboratorial tests. Hence, when new measurements are performed on samples of unknown status and the corresponding data is submitted to such an algorithm, it will be possible to predict whether the source individual is infected or not. This work presents a new methodology for rapid and precise label-free diagnosing of SARS-CoV-2 infection in clinical samples, which combines spectroscopic data acquisition and analysis via artificial intelligence algorithms. Our results show an accuracy of 85% for detection of SARS-CoV-2 in nasopharyngeal swab samples collected from asymptomatic patients or with mild symptoms, as well as an accuracy of 97% in tracheal aspirate samples collected from critically ill COVID-19 patients under mechanical ventilation. Moreover, the acquisition and processing of the information is fast, simple, and cheaper than traditional approaches, suggesting this methodology as a promising tool for biomedical diagnosis vis-à-vis the emerging and re-emerging viral SARS-CoV-2 variant threats in the future.

Ceccon Denny M, Amaral Paulo Henrique R, Andrade Lídia M, da Silva Maria I N, Andrade Luis A F, Moraes Thais F S, Bagno Flavia F, Rocha Raissa P, de Almeida Marques Daisymara Priscila, Ferreira Geovane Marques, Lourenço Alice Aparecida, Ribeiro Ágata Lopes, Coelho-Dos-Reis Jordana G A, da Fonseca Flavio G, Gonzalez J C

2023-Feb-28

Artificial intelligence, COVID-19, Label-free diagnosis, Machine learning, Optical spectroscopy

General General

Open Science Discovery of Potent Non-Covalent SARS-CoV-2 Main Protease Inhibitors

bioRxiv Preprint

The COVID-19 pandemic was a stark reminder that a barren global antiviral pipeline has grave humanitarian consequences. Pandemics could be prevented in principle by accessible, easily deployable broad-spectrum oral antivirals. Here we report the results of the COVID Moonshot, a fully open-science, crowd sourced, structure-enabled drug discovery campaign targeting the SARS-CoV-2 main protease. We discovered a novel chemical series that is differentiated from current Mpro inhibitors in that it maintains a new non-covalent, non-peptidic scaffold with nanomolar potency. Our approach leveraged crowdsourcing, high-throughput structural biology, machine learning, and exascale molecular simulations and high-throughput chemistry. In the process, we generated a detailed map of the structural plasticity of the SARS-CoV-2 main protease, extensive structure-activity relationships for multiple chemotypes, and a wealth of biochemical activity data. In a first for a structure-based drug discovery campaign, all compound designs (>18,000 designs), crystallographic data (>840 ligand-bound X-ray structures), assay data (>10,000 measurements), and synthesized molecules (>2,400 compounds) for this campaign were shared rapidly and openly, creating a rich open and IP-free knowledgebase for future anti-coronavirus drug discovery.

The COVID Moonshot Consortium, ; Achdout, H.; Aimon, A.; Alonzi, D. S.; Arbon, R.; Bar-David, E.; Barr, H.; Ben-Shmuel, A.; Bennett, J.; Bilenko, V. A.; Bilenko, V. A.; Boby, M. L.; Borden, B.; Boulet, P.; Bowman, G. R.; Brun, J.; Brwewitz, L.; BVNBS, S.; Calmiano, M.; Carbery, A.; Carney, D.; Cattermole, E.; Chang, E.; Chernyshenko, E.; Chodera, J. D.; Clyde, A.; Coffland, J. E.; Cohen, G.; Cole, J.; Contini, A.; Cox, L.; Croll, T. I.; Cvitkovic, M.; Dias, A.; Donckers, K.; Dotson, D. L.; Douangamath, A.; Duberstein, S.; Dudgeon, T.; Dunnett, L.; Eastman, P. K.; Erez, N.; Eyermann, C. J.; Fa

2023-03-02

Radiology Radiology

Large-Scale Domain-Specific Pretraining for Biomedical Vision-Language Processing

ArXiv Preprint

Contrastive pretraining on parallel image-text data has attained great success in vision-language processing (VLP), as exemplified by CLIP and related methods. However, prior explorations tend to focus on general domains in the web. Biomedical images and text are rather different, but publicly available datasets are small and skew toward chest X-ray, thus severely limiting progress. In this paper, we conducted by far the largest study on biomedical VLP, using 15 million figure-caption pairs extracted from biomedical research articles in PubMed Central. Our dataset (PMC-15M) is two orders of magnitude larger than existing biomedical image-text datasets such as MIMIC-CXR, and spans a diverse range of biomedical images. The standard CLIP method is suboptimal for the biomedical domain. We propose BiomedCLIP with domain-specific adaptations tailored to biomedical VLP. We conducted extensive experiments and ablation studies on standard biomedical imaging tasks from retrieval to classification to visual question-answering (VQA). BiomedCLIP established new state of the art in a wide range of standard datasets, substantially outperformed prior VLP approaches. Surprisingly, BiomedCLIP even outperformed radiology-specific state-of-the-art models such as BioViL on radiology-specific tasks such as RSNA pneumonia detection, thus highlighting the utility in large-scale pretraining across all biomedical image types. We will release our models at https://aka.ms/biomedclip to facilitate future research in biomedical VLP.

Sheng Zhang, Yanbo Xu, Naoto Usuyama, Jaspreet Bagga, Robert Tinn, Sam Preston, Rajesh Rao, Mu Wei, Naveen Valluri, Cliff Wong, Matthew P. Lungren, Tristan Naumann, Hoifung Poon

2023-03-02

Radiology Radiology

Deep learning-based decision forest for hereditary clear cell renal cell carcinoma segmentation on MRI.

In Medical physics ; h5-index 59.0

BACKGROUND : von Hippel-Lindau syndrome (VHL) is an autosomal dominant hereditary syndrome with an increased predisposition of developing numerous cysts and tumors, almost exclusively clear cell renal cell carcinoma (ccRCC). Considering the lifelong surveillance in such patients to monitor the disease, patients with VHL are preferentially imaged using MRI to eliminate radiation exposure.

PURPOSE : Segmentation of kidney and tumor structures on MRI in VHL patients is useful in lesion characterization (e.g. cyst vs. tumor), volumetric lesion analysis and tumor growth prediction. However, automated tasks such as ccRCC segmentation on MRI is sparsely studied. We develop segmentation methodology for ccRCC on T1 weighted pre-contrast, corticomedullary, nephrogenic and excretory contrast phase MRI.

METHODS : We applied a new neural network approach using a novel differentiable decision forest, called hinge forest (HF), to segment kidney parenchyma, cyst and ccRCC tumors in 117 images from 115 patients. This data set represented an unprecedented 504 ccRCCs with 1171 cystic lesions obtained at 5 different MRI scanners. The hinge forest architecture was compared with U-Net on 10 randomized splits with 75% used for training and 25% used for testing. Both methods were trained with Adam using default parameters (ɑ = 0.001, β1 = 0.9, β2 = 0.999) over 1000 epochs. We further demonstrated some interpretability of our HF method by exploiting decision tree structure.

RESULTS : The HF achieved an average kidney, cyst and tumor Dice similarity coefficient (DSC) of 0.75 ± 0.03, 0.44 ± 0.05, 0.53 ± 0.04 respectively while U-Net achieved an average kidney, cyst and tumor DSC of 0.78 ± 0.02, 0.41 ± 0.04, 0.46 ± 0.05 respectively. The HF significantly outperformed U-Net on tumors while U-Net significantly outperformed HF when segmenting kidney parenchymas (ɑ < 0.01).

CONCLUSIONS : For the task of ccRCC segmentation, the HF can offer better segmentation performance compared to the traditional U-Net architecture. The leaf maps can glean hints about deep learning features that might prove to be useful in other automated tasks such as tumor characterization.

Lay Nathan, Anari Pouria Yazdian, Chaurasia Aditi, Firouzabadi Mina Dehghani, Harmon Stephanie, Turkbey Evrim, Gautam Rabindra, Samimi Safa, Merino Maria J, Ball Mark W, Linehan W Marston, Turkbey Baris, Malayeri Ashkan

2023-Mar-01

MRI, deep learning, von Hippel-Lindau