From f7b9a51c47fb04648336d6574c8ce780b8e4d7d5 Mon Sep 17 00:00:00 2001
From: DevSoftTech Server <info.devsofttech@gmail.com>
Date: Tue, 31 Mar 2026 05:12:56 +0530
Subject: [PATCH] fix: sanitize noisy OCR output for OPD scan AI results

- add sanitizeExtractedScanText() to clean OCR text before analysis/storage
- strip non-ASCII/control noise that appears as mixed-script junk characters
- remove repeated junk tokens (e.g. ffffff, repeated words)
- add ECG-specific OCR normalization (KECG->ECG, Uniconfirmed->Unconfirmed, avt->aVL, VT->V1)
- apply sanitization to both OCR sources (document intelligence and vision fallback)
- keep fallback behavior when sanitization returns empty
---
 app/Controllers/Opd.php | 49 +++++++++++++++++++++++++++++++++++++++--
 1 file changed, 47 insertions(+), 2 deletions(-)

diff --git a/app/Controllers/Opd.php b/app/Controllers/Opd.php
index f3aab31..e78f75b 100644
--- a/app/Controllers/Opd.php
+++ b/app/Controllers/Opd.php
@@ -4584,18 +4584,63 @@ class Opd extends BaseController
 
         $docIntelText = $this->extractTextWithAzureDocumentIntelligence($filePath);
         if ($docIntelText !== '') {
-            return preg_replace('/\s+/', ' ', trim($docIntelText)) ?: trim($docIntelText);
+            $clean = $this->sanitizeExtractedScanText($docIntelText, $scanType);
+            return $clean !== '' ? $clean : (preg_replace('/\s+/', ' ', trim($docIntelText)) ?: trim($docIntelText));
         }
 
         $azurePrompt = 'Read and extract all visible clinical text from this ' . strtoupper($scanType) . ' medical image. Return plain text only. If unclear, return best readable lines.';
         $azureText = $this->callAzureOpenAiVision($filePath, $mimeType, $azurePrompt, 1800, false);
         if ($azureText !== '') {
-            return preg_replace('/\s+/', ' ', trim($azureText)) ?: trim($azureText);
+            $clean = $this->sanitizeExtractedScanText($azureText, $scanType);
+            return $clean !== '' ? $clean : (preg_replace('/\s+/', ' ', trim($azureText)) ?: trim($azureText));
         }
 
         return null;
     }
 
+    /**
+     * Clean OCR output so OPD UI stores readable clinical text only.
+     */
+    private function sanitizeExtractedScanText(string $text, string $scanType): string
+    {
+        $text = trim($text);
+        if ($text === '') {
+            return '';
+        }
+
+        // Remove control and non-ASCII characters that usually come from OCR noise.
+        $text = preg_replace('/[^\x09\x0A\x0D\x20-\x7E]/', ' ', $text) ?? $text;
+
+        // Drop repeated-char junk words like "ffffff".
+        $text = preg_replace('/\b([A-Za-z])\1{4,}\b/', ' ', $text) ?? $text;
+
+        // Collapse very long repeated words into one token.
+        $text = preg_replace('/\b([A-Za-z]{2,})\b(?:\s+\1\b){2,}/i', '$1', $text) ?? $text;
+
+        $isEcg = strtolower(trim($scanType)) === 'ecg' || stripos($text, 'ecg') !== false;
+        if ($isEcg) {
+            // Common OCR fixes for ECG templates.
+            $replacements = [
+                '/\bKECG\b/i' => 'ECG',
+                '/\bUniconfirmed\b/i' => 'Unconfirmed',
+                '/\bavt\b/i' => 'aVL',
+                '/\baVR\b/i' => 'aVR',
+                '/\baVF\b/i' => 'aVF',
+                '/\baVL\b/i' => 'aVL',
+                '/\bVT\b/' => 'V1',
+            ];
+
+            foreach ($replacements as $pattern => $replacement) {
+                $text = preg_replace($pattern, $replacement, $text) ?? $text;
+            }
+        }
+
+        // Normalize whitespace for compact storage/response.
+        $text = preg_replace('/\s+/', ' ', trim($text)) ?? trim($text);
+
+        return $text;
+    }
+
     /**
      * @return array{updated:bool,session_id:int,target_field:string}
      */