Background:The early diagnosis of rheumatic diseases improves their prognosis. However, patients take several months to reach the rheumatologist from the beginning of the first symptoms. Thermography is a safe and fast technique that captures the heat of an object through infrared photography. The inflammation of the joints causes an increase in temperature and, therefore, can be measured by thermography. Machine learning methods have shown that they are capable of analyzing medical images with an accuracy similar or superior to that of a healthcare professional.Objectives:Develop an algorithm that, based on thermographic images of hands and machine learning, differentiates healthy subjects from patients with rheumatoid arthritis (RA), psoriatic arthritis (PA), undifferentiated arthritis (UA) and arthritis of hands secondary to other diseases (SA).Methods:Multicenter observational study conducted in the rheumatology and radiology service of two hospitals. Patients with RA, PA, UA and SA who attended the followup visit and healthy subjects (companions and healthcare proffesionals) were recruited. In all cases, a thermal image of the hands was taken using a Flir One Pro or Thermal Expert TE-Q1 camera connected to the mobile and an ultrasound of both hands. The degree of synovial hypertrophy (SH) and power doppler (PD) was assessed for each joint (score from 0 to 3). Inflammation was defined as the presence of SH> 1 or PD> 0. Machine learning was used to classify patients with RA, PA, UA and SA with inflammation evidenced by ultrasound and healthy subjects from thermographic images. The evaluation of the classifier was performed by leave-one-out cross-validation and the area under the ROC curve (AUCROC) in those subjects whose thermal image was performed with the Thermal Expert TE-Q1 camera. The study was approved by the Clinical Ethics and Research Committee of the centers.Results:500 subjects were recruited from March 2018 to January 2020, of these 73 were excluded due to poor quality in the thermal image (moved or absence of temperature contrast between hand and background). Of the 427 subjects analyzed, 129 corresponded to healthy subjects, 138 to patients without evidence of inflammation and 160 to patients with inflammation evidenced by ultrasound (116 RA and 44 PA, UA or SA). Of these, 42% were taken using the Thermal Expert TE-Q1 camera. An AUCROC of 0.73 (p-value <0.01) was obtained for the healthy classifier vs RA and 0.72 (p-value <0.01) for the healthy classifier vs PA, UA and SA.Conclusion:A classification model has been developed capable of differentiating patients with RA, PA, UA and SA with evidence of inflammation from healthy subjects. These results open an opportunity to develop tools that facilitate early diagnosis.References:[1]Barhamain AS, Magliah RF, Shaheen MH, Munassar SF, Falemban AM, Alshareef MM, Almoallim HM. The journey of rheumatoid arthritis patients: a review of reported lag times from the onset of symptoms. Open Access Rheumatol. 2017 Jul 28;9:139-150. doi: 10.2147/OARRR.S138830. eCollection 2017. Review.[2]Lynch CJ, Liston C. New machine-learning technologies for computer-aided diagnosis. Nat Med. 2018 Sep;24(9):1304-1305. doi: 10.1038/s41591-018-0178-4.[3]Brenner M, Braun C, Oster M, Gulko PS. Thermal signature analysis as a novel method for evaluating inflammatory arthritis activity. Ann Rheum Dis. 2006 Mar;65(3):306-11.Disclosure of Interests:None declared
Forecasting the Environmental Parameters of Water Resources Using Machine Learning Methods
