JEVIKA, JEVIKA (2026) IMPLEMENTASI COMPUTER VISION UNTUK ANALISIS TEKANAN DARAH DAN SISTEM REKAM MEDIS ELEKTRONIK. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
cover.pdf Download (404kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
bab 1.pdf Restricted to Registered users only Download (171kB) |
|
|
Text (BAB II)
bab 2.pdf Restricted to Registered users only Download (722kB) |
|
|
Text (BAB III)
bab 3.pdf Restricted to Registered users only Download (847kB) |
|
|
Text (BAB IV)
bab 4.pdf Restricted to Registered users only Download (747kB) |
|
|
Text (BAB V)
bab 5.pdf Restricted to Registered users only Download (137kB) |
|
|
Text (DAFTAR PUSTAKA)
daftar pustaka.pdf Restricted to Registered users only Download (234kB) |
|
|
Text (LAMPIRAN)
lampiran.pdf Restricted to Registered users only Download (4MB) |
Abstract
This study developed a prototype blood-pressure history recording system integrated with face-recognition-based patient identification, transmission of digital blood pressure monitor readings through an ESP32, and exploratory analysis of facial color characteristics in the CIELAB color space. The face analysis was conducted on 21 videos collected across four acquisition sets: Office 1, Office 2, Field, and Home. Color features were extracted from the forehead, cheeks, nose, and entire face using MediaPipe Face Landmarker and were subsequently analyzed for their correlations with systolic blood pressure (SBP) and diastolic blood pressure (DBP) using Spearman correlation within each acquisition set. The patient identification module was developed using the InsightFace buffalo_s pipeline, which consists of SCRFD for face detection and ArcFace for embedding extraction. Face-recognition evaluation using 10-fold cross-validation achieved an accuracy of 94.07% on the LFW dataset and 98.22% on the CFP-FP dataset. Operational testing involving participants under normal conditions and simulated unconscious conditions showed that all participants were successfully identified, with no frames incorrectly assigned to another patient’s identity. The overall average identification time was 5.5 seconds under normal conditions and 11.2 seconds under simulated unconscious conditions. The system also successfully transmitted blood pressure readings from the digital monitor to the server, displayed the readings within the application, and stored the examination results in the corresponding patient profile. In the Office 1 set, the Fh_b_mean feature showed a strong and statistically significant positive correlation with SBP (ρ = 0.813; p = 0.002). These findings indicate a potential monotonic relationship between forehead color characteristics and systolic blood pressure under certain acquisition conditions. However, the CIELAB findings remain exploratory because of the limited sample size and cannot yet be used as a diagnostic tool for blood pressure assessment. Keywords: blood pressure, CIELAB, facial recognition, InsightFace, electronic medical records Penelitian ini mengembangkan prototipe sistem pencatatan riwayat tekanan darah yang terintegrasi dengan identifikasi pasien berbasis pengenalan wajah, pengiriman hasil pengukuran tensimeter digital melalui ESP32, dan analisis eksploratif karakteristik warna wajah dalam ruang CIELAB. Analisis wajah dilakukan terhadap 21 video dari empat set pengambilan, yaitu Kantor 1, Kantor 2, Lapangan, dan Rumah. Fitur warna dihitung pada area dahi, pipi, hidung, dan keseluruhan wajah menggunakan MediaPipe Face Landmarker, kemudian dianalisis korelasinya dengan tekanan darah sistolik (SBP) dan diastolik (DBP) menggunakan korelasi Spearman pada setiap set pengambilan. Modul identifikasi pasien dibangun menggunakan pipeline InsightFace buffalo_s yang terdiri atas SCRFD untuk deteksi wajah dan ArcFace untuk ekstraksi embedding. Evaluasi pengenalan wajah menggunakan 10-fold cross-validation menghasilkan akurasi sebesar 94,07% pada dataset LFW dan 98,22% pada CFP-FP. Pengujian operasional terhadap responden dalam kondisi normal dan simulasi tidak sadar menunjukkan bahwa seluruh responden berhasil diidentifikasi dan tidak ditemukan frame yang salah mengidentifikasi identitas pasien. Rata-rata waktu identifikasi secara keseluruhan adalah 5,5 detik pada kondisi normal dan 11,2 detik pada kondisi simulasi tidak sadar. Sistem juga berhasil mengirimkan data tekanan darah dari tensimeter ke server, menampilkannya pada sistem, dan menyimpan hasil pemeriksaan pada profil pasien yang sesuai. Pada set Kantor 1, fitur Fh_b_mean memiliki korelasi positif dan signifikan dengan SBP (ρ = 0,813; p = 0,002). Hasil tersebut menunjukkan adanya potensi hubungan monotonik antara karakteristik warna dahi dan tekanan darah sistolik pada kondisi pengambilan tertentu. Namun, temuan CIELAB masih bersifat eksploratif karena jumlah sampel yang terbatas dan belum dapat digunakan sebagai alat diagnosis tekanan darah. Kata Kunci : tekanan darah, CIELAB, pengenalan wajah, InsightFace, rekam medis elektronik
Actions (login required)
 |
View Item |
