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

General General

When will the mist clear? On the Interpretability of Machine Learning for Medical Applications: a survey

ArXiv Preprint

Artificial Intelligence is providing astonishing results, with medicine being one of its favourite playgrounds. In a few decades, computers may be capable of formulating diagnoses and choosing the correct treatment, while robots may perform surgical operations, and conversational agents could interact with patients as virtual coaches. Machine Learning and, in particular, Deep Neural Networks are behind this revolution. In this scenario, important decisions will be controlled by standalone machines that have learned predictive models from provided data. Among the most challenging targets of interest in medicine are cancer diagnosis and therapies but, to start this revolution, software tools need to be adapted to cover the new requirements. In this sense, learning tools are becoming a commodity in Python and Matlab libraries, just to name two, but to exploit all their possibilities, it is essential to fully understand how models are interpreted and which models are more interpretable than others. In this survey, we analyse current machine learning models, frameworks, databases and other related tools as applied to medicine - specifically, to cancer research - and we discuss their interpretability, performance and the necessary input data. From the evidence available, ANN, LR and SVM have been observed to be the preferred models. Besides, CNNs, supported by the rapid development of GPUs and tensor-oriented programming libraries, are gaining in importance. However, the interpretability of results by doctors is rarely considered which is a factor that needs to be improved. We therefore consider this study to be a timely contribution to the issue.

Antonio-Jesús Banegas-Luna, Jorge Peña-García, Adrian Iftene, Fiorella Guadagni, Patrizia Ferroni, Noemi Scarpato, Fabio Massimo Zanzotto, Andrés Bueno-Crespo, Horacio Pérez-Sánchez


Pathology Pathology

Mapping out the philosophical questions of AI and clinical practice in diagnosing and treating mental disorders.

In Journal of evaluation in clinical practice

How to classify the human condition? This is one of the main problems psychiatry has struggled with since the first diagnostic systems. The furore over the recent editions of the diagnostic systems DSM-5 and ICD-11 has evidenced it to still pose a wicked problem. Recent advances in techniques and methods of artificial intelligence and computing power which allows for the analysis of large data sets have been proposed as a possible solution for this and other problems in classification, diagnosing, and treating mental disorders. However, mental disorders contain some specific inherent features, which require critical consideration and analysis. The promises of AI for mental disorders are threatened by the unmeasurable aspects of mental disorders, and for this reason the use of AI may lead to ethically and practically undesirable consequences in its effective processing. We consider such novel and unique questions AI presents for mental health disorders in detail and evaluate potential novel, AI-specific, ethical implications.

Uusitalo Susanne, Tuominen Jarno, Arstila Valtteri


diagnosis, medical ethics, philosophy of medicine, progress

General General

Diagnostic accuracy of a novel third generation esophageal capsule as a noninvasive detection method for Barrett's Esophagus: A pilot study.

In Journal of gastroenterology and hepatology ; h5-index 51.0

BACKGROUND AND AIM : Previous 2 generations of esophageal capsule did not show adequate detection rates for Barrett's esophagus (BE). We assessed the diagnostic accuracy of a novel third generation capsule with an improved frame rate of 35 frames per second for the detection of BE in a pilot study.

METHODS : This was a blinded prospective pilot study conducted at a tertiary medical center. Patients with known BE (at least C0M>1) who presented for endoscopic surveillance (May to October 2017) were included. All patients underwent novel esophageal capsule (PillCamTM UGI; Medtronic) ingestion using the simplified ingestion protocol followed by standard high definition upper endoscopy (EGD). Capsule endoscopy findings were interpreted by examiners blinded to endoscopy results and compared to endoscopic findings (gold standard). Following completion of both tests, a subjective questionnaire was provided to all patients regarding their experience.

RESULTS : Twenty patients [95%males, mean age 66.3 (+/-7.9)years] with BE undergoing surveillance EGD were eligible. The mean BE length was 3.5 (+/-2.7)cm. Novel esophageal capsule detected BE in 75% patients when images were compared to endoscopy. Novel capsule detected BE in 82% patients when the BE length was >=2cm. The mean esophageal transit time was 0.59 sec. On a subjective questionnaire, all 20 patients reported novel capsule as being more convenient compared to EGD.

CONCLUSIONS : In this pilot, single center study, novel esophageal capsule was shown to be not ready for population screening of BE. Studies integrating artificial intelligence into improved quality novel esophageal capsule should be performed for BE screening.

Duvvuri Abhiram, Desai Madhav, Vennelaganti Sreekar, Higbee April, Gorrepati Venkat Subhash, Dasari Chandra, Chandrasekar Viveksandeep Thoguluva, Vennalaganti Prashanth, Kohli Divyanshoo, Sathyamurthy Anjana, Rai Tarun, Sharma Prateek


General General

Exploring the Potential of Artificial Intelligence and Machine Learning to Combat COVID-19 and Existing Opportunities for LMIC: A Scoping Review.

In Journal of primary care & community health

BACKGROUND : In the face of the current time-sensitive COVID-19 pandemic, the limited capacity of healthcare systems resulted in an emerging need to develop newer methods to control the spread of the pandemic. Artificial Intelligence (AI), and Machine Learning (ML) have a vast potential to exponentially optimize health care research. The use of AI-driven tools in LMIC can help in eradicating health inequalities and decrease the burden on health systems.

METHODS : The literature search for this Scoping review was conducted through the PubMed database using keywords: COVID-19, Artificial Intelligence (AI), Machine Learning (ML), and Low Middle-Income Countries (LMIC). Forty-three articles were identified and screened for eligibility and 13 were included in the final review. All the items of this Scoping review are reported using guidelines for PRISMA extension for scoping reviews (PRISMA-ScR).

RESULTS : Results were synthesized and reported under 4 themes. (a) The need of AI during this pandemic: AI can assist to increase the speed and accuracy of identification of cases and through data mining to deal with the health crisis efficiently, (b) Utility of AI in COVID-19 screening, contact tracing, and diagnosis: Efficacy for virus detection can a be increased by deploying the smart city data network using terminal tracking system along-with prediction of future outbreaks, (c) Use of AI in COVID-19 patient monitoring and drug development: A Deep learning system provides valuable information regarding protein structures associated with COVID-19 which could be utilized for vaccine formulation, and (d) AI beyond COVID-19 and opportunities for Low-Middle Income Countries (LMIC): There is a lack of financial, material, and human resources in LMIC, AI can minimize the workload on human labor and help in analyzing vast medical data, potentiating predictive and preventive healthcare.

CONCLUSION : AI-based tools can be a game-changer for diagnosis, treatment, and management of COVID-19 patients with the potential to reshape the future of healthcare in LMIC.

Naseem Maleeha, Akhund Ramsha, Arshad Hajra, Ibrahim Muhammad Talal

COVID-19, artificial intelligence, low middle-income countries, machine learning, pandemic

General General

Virtual organelle self-coding for fluorescence imaging via adversarial learning.

In Journal of biomedical optics

SIGNIFICANCE : Our study introduces an application of deep learning to virtually generate fluorescence images to reduce the burdens of cost and time from considerable effort in sample preparation related to chemical fixation and staining.

AIM : The objective of our work was to determine how successfully deep learning methods perform on fluorescence prediction that depends on structural and/or a functional relationship between input labels and output labels.

APPROACH : We present a virtual-fluorescence-staining method based on deep neural networks (VirFluoNet) to transform co-registered images of cells into subcellular compartment-specific molecular fluorescence labels in the same field-of-view. An algorithm based on conditional generative adversarial networks was developed and trained on microscopy datasets from breast-cancer and bone-osteosarcoma cell lines: MDA-MB-231 and U2OS, respectively. Several established performance metrics-the mean absolute error (MAE), peak-signal-to-noise ratio (PSNR), and structural-similarity-index (SSIM)-as well as a novel performance metric, the tolerance level, were measured and compared for the same algorithm and input data.

RESULTS : For the MDA-MB-231 cells, F-actin signal performed the fluorescent antibody staining of vinculin prediction better than phase-contrast as an input. For the U2OS cells, satisfactory metrics of performance were archieved in comparison with ground truth. MAE is <0.005, 0.017, 0.012; PSNR is >40  /  34  /  33  dB; and SSIM is >0.925  /  0.926  /  0.925 for 4',6-diamidino-2-phenylindole/hoechst, endoplasmic reticulum, and mitochondria prediction, respectively, from channels of nucleoli and cytoplasmic RNA, Golgi plasma membrane, and F-actin.

CONCLUSIONS : These findings contribute to the understanding of the utility and limitations of deep learning image-regression to predict fluorescence microscopy datasets of biological cells. We infer that predicted image labels must have either a structural and/or a functional relationship to input labels. Furthermore, the approach introduced here holds promise for modeling the internal spatial relationships between organelles and biomolecules within living cells, leading to detection and quantification of alterations from a standard training dataset.

Nguyen Thanh, Bui Vy, Thai Anh, Lam Van, Raub Christopher, Chang Lin-Ching, Nehmetallah Georges


artificial intelligence, fluorescence imaging, microscopy

Radiology Radiology

Artificial intelligence evaluating primary thoracic lesions has an overall lower error rate compared to veterinarians or veterinarians in conjunction with the artificial intelligence.

In Veterinary radiology & ultrasound : the official journal of the American College of Veterinary Radiology and the International Veterinary Radiology Association

To date, deep learning technologies have provided powerful decision support systems to radiologists in human medicine. The aims of this retrospective, exploratory study were to develop and describe an artificial intelligence able to screen thoracic radiographs for primary thoracic lesions in feline and canine patients. Three deep learning networks using three different pretraining strategies to predict 15 types of primary thoracic lesions were created (including tracheal collapse, left atrial enlargement, alveolar pattern, pneumothorax, and pulmonary mass). Upon completion of pretraining, the algorithms were provided with over 22 000 thoracic veterinary radiographs for specific training. All radiographs had a report created by a board-certified veterinary radiologist used as the gold standard. The performances of all three networks were compared to one another. An additional 120 radiographs were then evaluated by three types of observers: the best performing network, veterinarians, and veterinarians aided by the network. The error rates for each of the observers was calculated as an overall and for the 15 labels and were compared using a McNemar's test. The overall error rate of the network was significantly better than the overall error rate of the veterinarians or the veterinarians aided by the network (10.7% vs 16.8% vs17.2%, P = .001). The network's error rate was significantly better to detect cardiac enlargement and for bronchial pattern. The current network only provides help in detecting various lesion types and does not provide a diagnosis. Based on its overall very good performance, this could be used as an aid to general practitioners while waiting for the radiologist's report.

Boissady Emilie, de La Comble Alois, Zhu Xiaojuan, Hespel Adrien-Maxence


computer vision-based decision support system, convolutional neural networks, deep learning, small animal thoracic radiology