Artificial Intelligence Algorithm with SVM Classification using Dermascopic Images for Melanoma Diagnosis

Of all suspicious pigmented skin lesions considered for analysis, a large portion is often benign. The pressure of pathology services and secondary care must be reduced throughout the patient trials using modern techniques for improving the melanoma diagnosis accuracy. Dermoscopic images obtained from digital single-lens reflex (DSLR) cameras, smartphones and a lightweight USB camera are compared using artificial intelligence (AI) algorithm for determining the accuracy of melanoma identification. Datasets are obtained from thousand test samples undergoing plastic surgery. The diagnostic trial is masked, single arm and multicentered. The controlled and suspicious skin lesions as well as the suspicious pigmented skin lesion are captured on the aforementioned cameras while scheduling for biopsy. The possibility of melanoma is assessed using deep learning (DL) techniques on the pigmented skin lesions seen in the dermascopic images for identifying melanoma. For this purpose, we train a deterministic AI algorithm based on malignancy recognition by deep ensemble and inputs from clinicians. The histopathology diagnosis is used as a standard criterion for determining the specialist assessment, algorithmic specificity, sensitivity and the area under the receiver operating characteristic curve (AUROC).

Download Full-text

Stress testing reveals gaps in clinic readiness of image-based diagnostic artificial intelligence models

npj Digital Medicine ◽

10.1038/s41746-020-00380-6 ◽

2021 ◽

Vol 4 (1) ◽

Author(s):

Albert T. Young ◽

Kristen Fernandez ◽

Jacob Pfau ◽

Rasika Reddy ◽

Nhat Anh Cao ◽

...

Keyword(s):

Artificial Intelligence ◽

Real World ◽

Stress Testing ◽

Operating Characteristic ◽

Characteristic Curve ◽

Skin Lesions ◽

Stress Tests ◽

Performance Metric ◽

Intelligence Models ◽

Artificial Intelligence Models

AbstractArtificial intelligence models match or exceed dermatologists in melanoma image classification. Less is known about their robustness against real-world variations, and clinicians may incorrectly assume that a model with an acceptable area under the receiver operating characteristic curve or related performance metric is ready for clinical use. Here, we systematically assessed the performance of dermatologist-level convolutional neural networks (CNNs) on real-world non-curated images by applying computational “stress tests”. Our goal was to create a proxy environment in which to comprehensively test the generalizability of off-the-shelf CNNs developed without training or evaluation protocols specific to individual clinics. We found inconsistent predictions on images captured repeatedly in the same setting or subjected to simple transformations (e.g., rotation). Such transformations resulted in false positive or negative predictions for 6.5–22% of skin lesions across test datasets. Our findings indicate that models meeting conventionally reported metrics need further validation with computational stress tests to assess clinic readiness.

Download Full-text

Identification of Skin Lesions by Using Single-Step Multiframe Detector

Journal of Clinical Medicine ◽

10.3390/jcm10010144 ◽

2021 ◽

Vol 10 (1) ◽

pp. 144

Author(s):

Yu-Ping Hsiao ◽

Chih-Wei Chiu ◽

Chih-Wei Lu ◽

Hong Thai Nguyen ◽

Yu Sheng Tseng ◽

...

Keyword(s):

Ground Truth ◽

Skin Lesions ◽

Single Step ◽

Single Shot ◽

Accuracy Rate ◽

Tissue Slices ◽

Image Detection ◽

Sensitivity Rate ◽

Diagnosis Accuracy ◽

Artificial Intelligence Algorithm

An artificial intelligence algorithm to detect mycosis fungoides (MF), psoriasis (PSO), and atopic dermatitis (AD) is demonstrated. Results showed that 10 s was consumed by the single shot multibox detector (SSD) model to analyze 292 test images, among which 273 images were correctly detected. Verification of ground truth samples of this research come from pathological tissue slices and OCT analysis. The SSD diagnosis accuracy rate was 93%. The sensitivity values of the SSD model in diagnosing the skin lesions according to the symptoms of PSO, AD, MF, and normal were 96%, 80%, 94%, and 95%, and the corresponding precision were 96%, 86%, 98%, and 90%. The highest sensitivity rate was found in MF probably because of the spread of cancer cells in the skin and relatively large lesions of MF. Many differences were found in the accuracy between AD and the other diseases. The collected AD images were all in the elbow or arm and other joints, the area with AD was small, and the features were not obvious. Hence, the proposed SSD could be used to identify the four diseases by using skin image detection, but the diagnosis of AD was relatively poor.

Download Full-text

Validation of an artificial intelligence algorithm applied to a metabolic substrate analysis of urine for detection of urothelial cancer.

Journal of Clinical Oncology ◽

10.1200/jco.2019.37.15_suppl.e16008 ◽

2019 ◽

Vol 37 (15_suppl) ◽

pp. e16008-e16008

Author(s):

Nikola Kaludov ◽

Mohummad Minhaj Siddiqui ◽

Max Kates ◽

Hemantkumar Tripathi ◽

Amatul Nasir Salma ◽

...

Keyword(s):

Artificial Intelligence ◽

Predictive Value ◽

Urothelial Cancer ◽

Profile Analysis ◽

Characteristic Curve ◽

Area Under The Curve ◽

Data Points ◽

Artificial Intelligence Algorithm ◽

Insight Into ◽

Made In

e16008 Background: Urine tests such as urine cytology are commonly used for the diagnosis and monitoring of urothelial cancer. These tests are often limited by issues related to sensitivity or specificity. It is well known that derangement of cellular metabolism is one of the hallmarks of carcinogenesis. As urothelial cancer is in constant contact with urine, we hypothesize that metabolite composition in the urine may provide insight into possible urothelial cancer presence in the urinary tract. In this study, we evaluated a metabolomics based urine test for the detection of urothelial cancer. Methods: In this prospective, multi-institutional IRB approved study, urine samples were collected from a total of 57 urothelial cancer patients and non-urothelial cancer controls. Gas chromatography profiles of urine small molecule metabolites were generated to yield over 2400 data points of metabolite peaks and troughs for every urine sample. A machine-learning based algorithm (Abilis Life Sciences) was constructed to predict urothelial cancer versus non-cancer controls through analysis of peaks and trough patterns of urine metabolomics profiles. Predictions were made in a blinded fashion and descriptive statistics of test sensitivity and specificity were generated. Results: The urine metabolite composition of 57 patients were analyzed and urothelial cancer predictions were generated. The test demonstrated an overall accuracy of 89.5% (51 out of 57 cases correctly predicted). The sensitivity of the test was 97.1% (34 out of 35) and specificity was 77.3% (17 out of 22). The Positive Predictive Value is 87.2%, while the Negative Predictive Value is 94.4%. The area under the curve for the receiver operating characteristic curve was 0.87. Conclusions: Urine based metabolic profile analysis using artificial intelligence algorithms is a promising potential diagnostic test for detection of urothelial cancer. Further testing is ongoing to increase robustness of the validation.

Download Full-text

DEEP LEARNING BASED SKIN LESIONS DIAGNOSIS

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ◽

10.5194/isprs-archives-xlii-2-w12-81-2019 ◽

2019 ◽

Vol XLII-2/W12 ◽

pp. 81-85 ◽

Cited By ~ 1

Author(s):

D. A. Gavrilov ◽

N. N. Shchelkunov ◽

A. V. Melerzanov

Keyword(s):

Neural Network ◽

Classification Accuracy ◽

Intelligent Systems ◽

State Of The Art ◽

Skin Lesions ◽

Deep Convolutional Neural Networks ◽

Melanoma Diagnosis ◽

Malignant Skin ◽

Diagnosis Accuracy ◽

Visual Diagnosis

<p><strong>Abstract.</strong> Melanoma is one of the most virulent lesions of human’s skin. The visual diagnosis accuracy of melanoma directly depends on the doctor’s qualification and specialization. State-of-the-art solutions in the field of image processing and machine learning allows to create intelligent systems based on artificial convolutional neural network exceeding human’s rates in the field of object classification, including the case of malignant skin lesions. This paper presents an algorithm for the early melanoma diagnosis based on artificial deep convolutional neural networks. The algorithm proposed allows to reach the classification accuracy of melanoma at least 91%.</p>

Download Full-text

Deep Neural Frameworks Improve the Accuracy of General Practitioners in the Classification of Pigmented Skin Lesions

Diagnostics ◽

10.3390/diagnostics10110969 ◽

2020 ◽

Vol 10 (11) ◽

pp. 969

Author(s):

Maximiliano Lucius ◽

Jorge De All ◽

José Antonio De All ◽

Martín Belvisi ◽

Luciana Radizza ◽

...

Keyword(s):

General Practitioners ◽

Skin Lesions ◽

Image Resolution ◽

Aggregated Data ◽

Skin Imaging ◽

High Image Resolution ◽

Random Dataset ◽

Algorithmic Framework ◽

Diagnosis Accuracy ◽

Pigmented Skin Lesions

This study evaluated whether deep learning frameworks trained in large datasets can help non-dermatologist physicians improve their accuracy in categorizing the seven most common pigmented skin lesions. Open-source skin images were downloaded from the International Skin Imaging Collaboration (ISIC) archive. Different deep neural networks (DNNs) (n = 8) were trained based on a random dataset constituted of 8015 images. A test set of 2003 images was used to assess the classifiers’ performance at low (300 × 224 RGB) and high (600 × 450 RGB) image resolution and aggregated data (age, sex and lesion localization). We also organized two different contests to compare the DNN performance to that of general practitioners by means of unassisted image observation. Both at low and high image resolution, the DNN framework differentiated dermatological images with appreciable performance. In all cases, the accuracy was improved when adding clinical data to the framework. Finally, the least accurate DNN outperformed general practitioners. The physician’s accuracy was statistically improved when allowed to use the output of this algorithmic framework as guidance. DNNs are proven to be high performers as skin lesion classifiers and can improve general practitioner diagnosis accuracy in a routine clinical scenario.

Download Full-text

Automatic Imaging System With Decision Support for Inspection of Pigmented Skin Lesions and Melanoma Diagnosis

IEEE Journal of Selected Topics in Signal Processing ◽

10.1109/jstsp.2008.2011156 ◽

2009 ◽

Vol 3 (1) ◽

pp. 14-25 ◽

Cited By ~ 99

Author(s):

Jose Fernandez Alcon ◽

Calina Ciuhu ◽

Warner ten Kate ◽

Adrienne Heinrich ◽

Natallia Uzunbajakava ◽

...

Keyword(s):

Decision Support ◽

Imaging System ◽

Skin Lesions ◽

Melanoma Diagnosis ◽

Pigmented Skin Lesions

Download Full-text

An artificial intelligence algorithm for analyzing acetaminophen-associated toxic hepatitis

Human & Experimental Toxicology ◽

10.1177/09603271211014587 ◽

2021 ◽

pp. 096032712110145

Author(s):

J-S Yen ◽

C-C Hu ◽

W-H Huang ◽

C-W Hsu ◽

T-H Yen ◽

...

Keyword(s):

Artificial Intelligence ◽

Platelet Count ◽

Prothrombin Time ◽

Operating Characteristic ◽

Toxic Hepatitis ◽

Characteristic Curve ◽

Chang Gung Memorial Hospital ◽

Algorithm Development ◽

Intelligence Algorithm ◽

Artificial Intelligence Algorithm

Introduction: Very little artificial intelligence (AI) work has been performed to investigate acetaminophen-associated hepatotoxicity. The objective of this study was to develop an AI algorithm for analyzing weighted features for toxic hepatitis after acetaminophen poisoning. Methods: The medical records of 187 patients with acetaminophen poisoning treated at Chang Gung Memorial Hospital were reviewed. Patients were sorted into two groups according to their status of toxic hepatitis. A total of 40 clinical and laboratory features recorded on the first day of admission were selected for algorithm development. The random forest classifier (RFC) and logistic regression (LR) were used for artificial intelligence algorithm development. Results: The RFC-based AI model achieved the following results: accuracy = 92.5 ± 2.6%; sensitivity = 100%; specificity = 60%; precision = 92.3 ± 3.4%; and F1 = 96.0 ± 1.8%. The area under the receiver operating characteristic curve (AUROC) was approximately 0.98. The LR-based AI model achieved the following results: accuracy = 92.00 ± 2.9%; sensitivity = 100%; specificity = 20%; precision = 92.8 ± 3.4%; recall = 98.8 ± 3.4%; and F1 = 95.6 ± 1.5%. The AUROC was approximately 0.68. The weighted features were calculated, and the 10 most important weighted features for toxic hepatitis were aspartate aminotransferase (ALT), prothrombin time, alanine aminotransferase (AST), time to hospital, platelet count, lymphocyte count, albumin, total bilirubin, body temperature and acetaminophen level. Conclusion: The top five weighted features for acetaminophen-associated toxic hepatitis were ALT, prothrombin time, AST, time to hospital and platelet count.

Download Full-text